Short answer: what should a new website do first?
Build for people and normal search first, then make that content easy for answer engines to retrieve. Get the technical basics right, publish a small number of genuinely useful pages around one topic, put a real expert behind them, make important facts easy to quote, and earn independent mentions off-site.
There is no special Google "GEO schema" or shortcut. Google says its generative search features are rooted in core Search systems, while ChatGPT Search has its own search crawler. Treat AEO/GEO as an extension of strong SEO — not a replacement for it.
Third-party statistics are attributed to their original sources. Where a number comes from IndexCraft's own research (n=89 sites, January–December 2025), it is labelled as internal and the methodology is available on request. First-party observations are kept separate from outside research so you can tell evidence from experience. Last reviewed and updated on October 6, 2026.
If you launched a website in 2021, the game was mostly: rank on page one, get clicked. In 2026, discovery is spread across classic results, AI Overviews, AI Mode, ChatGPT Search, Copilot and Perplexity. Ranking still matters, but the bigger goal is to become a source that search systems can understand, retrieve and trust.
For new sites, this cuts both ways. Yes, you start with no domain authority, no backlinks, and no ranking history — that gap is real. You do start with less external reputation and fewer links, but domain age itself is not the optimisation target. A focused new site can still earn visibility when its pages are crawlable, specific, well-sourced and genuinely better for a narrow task than the alternatives.
I've been tracking AI citation patterns across new site launches since Google AI Overviews launched in May 2024 — logging which queries each site appeared in across Perplexity, Google AI Overviews, and ChatGPT Search, and noting the exact date each first citation appeared.
The finding that comes through most consistently: in this tracked cohort, sites that launched with answer-first content, named authors and a clear crawlable information architecture reached their first Perplexity citation in an average of 19 days. Sites that launched homepage-first and retrofitted useful content later averaged 74 days. That is an internal observation, not proof that any one element caused the difference.
The clearest side-by-side I tracked was two B2B software sites targeting overlapping topics that launched within about two weeks of each other. One had a proper pillar page and six cluster articles ready on day one, all with schema in place. The other launched with a homepage, an About page, and three broad blog posts with no particular structure.
At the 90-day mark, the structured site had 21 active Perplexity citations. The other had none. The second site eventually caught up — but it had seven months of catch-up work to do in a space where the other site had already established citation history. That head start compounds faster than most people expect when they're planning a launch.
What follows is a practical roadmap for building AEO and GEO from zero: a phased 12-month plan, separate chapters for B2B and B2C, and tactical guidance drawn from auditing 47 new-site launches. Bookmark it and come back to it — the priorities shift as your site matures.
1. AEO vs GEO: Understanding the Distinction
Most people use AEO and GEO as synonyms. They are not quite the same thing, and the distinction matters in practice.
AEO (Answer Engine Optimisation)
The broader practice of structuring content so it can be extracted as a direct answer by any automated system — featured snippets, voice assistants, zero-click results, and AI chatbots. It has been evolving since featured snippets launched in 2014.
GEO (Generative Engine Optimisation)
More specific: it focuses on being cited by large language model-powered search platforms — Google AI Overviews, ChatGPT Search, and Perplexity AI. AEO is the parent discipline; GEO adds topical authority architecture, entity recognition, and LLM-specific citation signals on top of AEO foundations.
54.5%
of all Google AI Overview citations now overlap with organically-ranking pages — up from 32.3% in May 2024, a gain of 22.3 percentage points in 16 months. The convergence is strongest in YMYL categories: Healthcare (68–75% overlap) and Education (53.2 pp increase). E-commerce remains the outlier, with virtually flat overlap, suggesting very different source-selection behaviour by industry.
Don't line this up against Ahrefs' 38% as if it were one trend. BrightEdge's 54.5% counts citations that overlap with pages ranking organically at all, while its separate top-10-only figure is far lower, at roughly 17%. Different vendors, time windows and definitions of "overlap" produce different numbers.
Source: BrightEdge Generative Parser™, 16-month AI Overview citation overlap study, May 2024–September 2025For a new site, the takeaway is: get your AEO signals right first — proper structure, direct answers, schema, clear authorship — because those are what GEO layers on top of. Starting both at once is the most efficient route, and honestly the only sensible one when you have nothing yet.
2. The Reality for New Websites: What You're Up Against
New websites face some real, concrete obstacles here. Worth knowing what they are before you try to work around them.
The authority gap
Established brands do start with an advantage: more independent mentions, stronger pages, broader topical coverage and a longer history of being discovered. But "domain authority" is a third-party metric, not a Google ranking factor. For a new site, the practical gap to close is reputation and evidence — especially relevant off-site mentions, useful links, strong pages and clear entity information.
Brand mentions matter
In Ahrefs' 75,000-brand study, branded web mentions correlated more strongly with AI visibility than backlink count or Domain Rating. Their later cross-platform analysis found branded web mentions at roughly 0.66–0.71 correlation across ChatGPT, AI Mode and AI Overviews. Treat this as directional evidence, not a causal ranking formula.
Source: Ahrefs, “Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews” (December 2025)One pattern holds up whenever I look at low-authority sites that do get cited: they almost never got there by competing broadly. They are the most detailed source on something narrow enough that larger sites haven't gone deep on it.
I tracked a site covering a fairly specialised compliance sub-topic within one industry vertical — brand new domain, no backlinks, zero domain authority by any standard measure. Within about five weeks of launch, it was appearing in Perplexity for its core queries.
Not because it outranked anything — it had no rankings to speak of — but because every established site touching the broader category had only a paragraph or two on this specific angle, while this site had three or four articles going three levels deeper.
I checked their 14 target queries in Perplexity every two weeks for four months. By month three, they had citations on 9 of the 14. The high-authority broad-topic sites were never cited for those specific queries across any of my checks.
The lesson is the same one I keep coming back to with new sites: you cannot win a head-to-head against established domains on their terrain. You can win on yours.
The indexation lag
For Google AI features, a page must be indexed and eligible to appear in Search with a snippet. Other AI products have their own retrieval systems and crawler controls. Submit your XML sitemap to Google Search Console and Bing Webmaster Tools at launch, then verify that priority pages are actually crawlable and indexed instead of assuming submission guarantees discovery.
Indexability comes first
Google's current guidance is simple: there are no additional technical requirements for AI Overviews or AI Mode beyond being eligible for Google Search. Get crawling, indexing, snippets, internal links and page experience right before chasing AI-specific tactics.
Source: Google Search Central, “AI features and your website”The entity recognition gap
Language models understand the world through entities — named organisations, people, brands, concepts. A new brand with zero external mentions is essentially invisible to those systems; it has no entity. Every time your brand name appears on another site — even without a link — that is an entity signal. In Ahrefs' study, branded web mentions showed the strongest correlation with AI visibility, ahead of backlink count and Domain Rating. That is correlation, not proof of cause, but it is a good reason to start earning mentions early. The semantic SEO and entity guide covers the longer version of how entities work.
What is a Knowledge Graph entity, and why does it matter for new sites?
Google's Knowledge Graph is the database of real-world entities (people, organisations, products, concepts) that Google uses to interpret search queries and select trusted sources for AI Overviews. Entity consistency helps machines connect your organisation, people, products and external references. You do not need a visible Knowledge Panel before you can rank or be cited, but consistent names, author identities and external corroboration reduce ambiguity.
You build entity recognition through: consistent brand naming across all web properties, structured Organization schema with sameAs references, external brand mentions in publications and directories, and — for individuals — named author pages with verifiable credentials. Entity-building is not a one-time task; it compounds with every new external mention.
3. Building the Foundation: What to Set Up Before Publishing
Before the first article goes live, a few things need to be in place. Not aspirational extras — prerequisites. Skip these and the tactics later in this guide will not work properly. The technical SEO guide covers each of them in more depth.
Submit your XML sitemap to both at launch. Google Search Console now includes a dedicated Generative AI performance report for AI Overviews and AI Mode. Bing Webmaster Tools has an AI Performance report showing citations, cited pages and grounding-query insights across supported Microsoft AI experiences. Neither tool guarantees inclusion; they give you the best first-party visibility into discovery and citation activity.
Use HTTPS from day one. It is a basic security and user-trust requirement, and Google has long used HTTPS as a lightweight ranking signal. More importantly, browsers and users should never have to choose between safety warnings and your content. Most modern hosting platforms include automated certificates.
Decide separately what you want indexed for search and what you want used for model training. For ChatGPT Search discovery, OpenAI says not to block OAI-SearchBot. GPTBot is a separate crawler used for potential model training. Google Search uses Googlebot; Google-Extended is a control for certain Gemini/Vertex AI uses and does not control ranking or inclusion in Google Search. If Perplexity visibility matters to you, allow its search crawler as well.
User-agent: *
Allow: /
# ChatGPT Search discovery
User-agent: OAI-SearchBot
Allow: /
# Perplexity search discovery
User-agent: PerplexityBot
Allow: /
# Training controls are separate decisions:
# User-agent: GPTBot
# Disallow: /
# User-agent: Google-Extended
# Disallow: /Keep normal search crawlers such as Googlebot and Bingbot unblocked through your site-wide rule. Review each vendor's documentation before changing training permissions.
Keep URLs short, descriptive and stable. A slug like /what-is-supply-chain-management is easier for people to understand and maintain than an opaque ID. Avoid dates in evergreen URLs when the date is not part of the resource identity; the main reason is maintainability and longevity, not a special AI-citation penalty.
Use sensible headings, paragraphs, lists, tables and semantic elements where they improve the document. Google explicitly says perfectly semantic HTML is not required, but clear structure helps users — including assistive technologies — navigate the page and makes important text easier to understand. Do not turn semantic markup into a checklist of fake ranking signals.
Source: Google Search Central, “Optimizing your website for generative AI features on Google Search”E-E-A-T assessment starts with "who is behind this content?" If your site has no About page, no named authors, and no identifiable organisation, it fails the most basic trust check. Publish a detailed About page and individual author bio pages before you publish your first article. For topics where people need to judge expertise or accountability, make authorship and editorial responsibility easy to verify. Quality Rater Guidelines are evaluation guidance, not a direct ranking checklist, so use them to improve transparency rather than to manufacture credentials.
Add a JSON-LD Organization schema block to your homepage that declares your brand name, URL, logo, and contact information — including your sameAs references to your web profiles and directory listings. This is the foundation of entity recognition: it tells search systems and LLMs that your brand is a real, identifiable organisation with a verifiable external presence.
4. Content Architecture for AI-First Websites
Before you write anything, decide how your content will be organised. What topics will you cover? How do they connect? How will pages link to each other? Getting this right on day one is much easier than untangling a disorganised site at month eight — and it is the single structural factor I see separating sites that build AI citation authority quickly from sites that keep wondering why nothing is working.
The topic cluster model (explained simply)
The structure works like this: one pillar page covers a broad topic comprehensively. Around it sit cluster pages — focused articles that each go deep on a specific sub-topic. Cluster pages link back to the pillar; the pillar links out to each cluster. Simple, but it matters. The topical authority and pillar pages guide goes deeper on the model, and the internal linking strategy guide shows how to wire it together.
The effect is cumulative. For traditional SEO, internal linking concentrates topical relevance around your most important page. For AI search, the whole cluster signals depth of coverage — making your site look like an authority on the topic rather than a site with one decent article and nothing behind it.
Think in topics, not keyword clones
Google's current generative-search documentation confirms that AI features use query fan-out: the system can run related searches to gather supporting information. That makes complete topic coverage useful — but Google also warns against creating a separate page for every query variation. Build a tight cluster because users need the depth, not because a fixed article count unlocks an AI score.
Source: Google Search Central, “Optimizing your website for generative AI features on Google Search”How many topic clusters should a new site build?
For a new site, focus on one or two clear topic areas in the first 90 days. The exact article count is less important than whether each page solves a distinct user task. A pillar plus a handful of genuinely different supporting pages is usually a better starting point than dozens of near-duplicate keyword variations.
I audited a content site at around the six-month mark that had done the work — published consistently, decent writing, reasonable topic selection on each individual piece — but AI citations were essentially flat. Under half a citation per month across all platforms.
When I mapped the content, the problem was obvious: 68 articles spread across five loosely related topics with no pillar structure connecting any of them. It looked like a site that had been publishing whatever seemed interesting at the time rather than building depth on anything specific.
We pulled the architecture back to two tight topic clusters — consolidating and redirecting the thinner articles, strengthening the ones worth keeping, rebuilding internal linking around the new structure. The actual content barely changed.
By month nine, citation activity across Perplexity and Google AI Overviews had moved from that 0.4 per month baseline to around 3.8 per month. Same words, essentially, just organised in a way that reads as depth rather than breadth. That architecture decision is probably the single highest-leverage choice a new site makes in its first year.
Question mapping: the content brief framework
Before you write anything, spend time on question research. Pull from Google's "People Also Ask" boxes, Reddit threads, Quora, LinkedIn comments, and AnswerThePublic. For each article you plan: know the main question it answers, three to five secondary questions it should also cover, and the one-sentence direct answer you will put in the opening paragraph. For the research step itself, the conversational keyword research guide walks through a method.
5. The 12-Month Phased Action Plan
The order matters, but the article count does not. The safest plan is to prove one topic, one audience and one content system before you scale. Google specifically warns against producing lots of query-variation pages just to manipulate Search or AI responses, so use the timeline below as a quality sequence rather than a publishing quota.
D0 — Pre-Launch: Make the Site Discoverable and Trustworthy
- Set up Google Search Console and Bing Webmaster Tools; submit the XML sitemap
- Check robots.txt, CDN rules and meta robots; allow Googlebot and the search crawlers you actually want
- Allow OAI-SearchBot if you want your public pages eligible for ChatGPT Search discovery
- Publish a real About page, editorial policy and named author profiles
- Implement Organization, Article and Breadcrumb markup where the visible content supports it
- Record a mobile Lighthouse baseline and watch field Core Web Vitals once traffic is available (the Core Web Vitals guide shows how)
M1 — Month 1: Build One Useful Topic Cluster
Start with the questions your ideal visitor must understand before they can evaluate a solution. Publish only what you can make meaningfully better than the current results.
- Publish one strong pillar or hub page with clear navigation to supporting resources
- Publish 4–6 supporting pages that solve genuinely different tasks — definitions, decisions, setup questions or common problems
- Lead each page with the answer, then add evidence, examples, caveats and next steps
- Use a named author and link to a detailed bio; show publish/updated dates accurately
- Request indexing only for a few high-priority URLs when necessary; let the sitemap and internal links do the routine work
M2 — Month 2: Add Decision and Action Content
- Publish comparison content with transparent criteria, not affiliate-style “best” lists with no methodology
- Publish step-by-step guides using real screenshots, outcomes or first-hand notes where possible
- Add concise FAQs only where they answer real follow-up questions; do not force FAQ schema for an imagined Google boost
- Create your first source-worthy asset: benchmark, template, calculator, teardown, dataset or small original study
- Begin outreach to relevant publications and communities where the research is genuinely useful
M3 — Month 3: Build External Corroboration
- Publish your first original-data page with a transparent methodology and limitations
- Standardise brand names, author names and profile information across the web
- Earn a small number of relevant editorial mentions instead of chasing bulk directory links
- Start a fixed prompt/query set for monthly AI visibility checks so results are comparable
- Review Google Search Console's Generative AI report and Bing Webmaster Tools AI Performance data when available
M4–6 — Months 4–6: Deepen What Is Working
Ahrefs' March 2026 study found that 37.9% of URLs cited in AI Overviews also appeared in the first ten SERP blocks. That is strong enough to keep organic SEO central, but low enough to show that query fan-out can bring in additional sources. An earlier Ahrefs study of 1.9M citations (mid-2025) found 76%, so the top-10 share has fallen. The samples differ, so treat the drop as directional.
- Improve pages already earning impressions before opening a large second cluster
- Add missing examples, comparison criteria, original visuals and expert commentary
- Build a second cluster only when the first has clear audience demand and internal-link logic
- Refresh facts that changed; keep URLs stable unless there is a real information-architecture reason to move them
- Measure conversions and assisted discovery, not citation counts in isolation
M7–9 — Months 7–9: Improve Extraction, Experience and Reputation
- Rewrite vague introductions so the first useful answer appears early without sounding robotic
- Use tables for true comparisons, ordered lists for real procedures, and prose for nuance
- Add original images or video where the task is visual; optimise filenames, alt text, dimensions and loading
- Close mention gaps: identify credible publications, communities and videos discussing competitors but not your brand
- Consolidate overlapping pages instead of maintaining several weak variants
M10–12 — Months 10–12: Refresh Winners and Scale Carefully
- Refresh content where the underlying facts, products, screenshots or recommendations changed
- Use Search Console and Bing data to identify topics where you appear but underperform
- Study the exact passages that get cited, but preserve natural writing instead of turning every paragraph into a “chunk”
- Expand into adjacent topics only where you can add expert or first-hand value
- Run a technical, structured-data, accessibility and Core Web Vitals review before the next scale phase
6. B2B AEO/GEO Strategy: The Complete Playbook
B2B buyers use AI search differently from consumers. They are mid-research — evaluating vendors, building a business case, validating a technical claim. Their queries are longer and more specific. The content they need is deeper and more evidence-heavy. A blog post that would work well for a B2C audience will feel thin to a procurement manager who has already done three rounds of AI research before they reach your site.
Evidence beats generic volume
B2B AI research tends to branch into implementation, risk, cost, integration and vendor-comparison questions. Build content that helps a buyer verify a claim and make a decision: real implementation details, named experts, technical constraints, case evidence and transparent comparison criteria.
Strategy note: optimise for the full research journey rather than one headline statistic.B2B content priorities for AI citation
Content types that drive B2B AI citations
Industry-specific definitional and explainer articles · Technical comparison guides (your product vs competitors) · Original industry research and benchmark reports · Case studies with specific, named metrics and client outcomes · ROI calculation guides and frameworks · Glossaries of industry terminology · Regulatory and compliance explainers · Integration and technical capability documentation
Queries your B2B buyers use in AI search
"What is [technology/process] and how does it work?" · "[Product A] vs [Product B] for [specific use case]" · "How to implement [solution] at enterprise scale" · "What does [industry term] mean in [context]?" · "Best [software category] for [company size/industry]" · "How to build a business case for [investment]" · "What are the risks of [approach/technology]?" · "[Regulation] compliance requirements for [industry]"
B2B entity and author authority
In B2B, authorship is not a formality. A post from "the TechSolutions team" carries a fraction of the weight of an article attributed to "Sarah Chen, VP of Engineering, 12 years in enterprise cloud infrastructure." B2B readers — and the AI systems they use — are checking that the person behind the content actually knows what they are talking about.
I've run variations of this authorship test across several client sites now, and the result is consistent enough that named attribution comes up in almost every B2B content briefing I give.
The setup is usually the same: two content clusters on the same site, covering similar topics, one bylined to a generic team or editorial credit, one attributed to a named person with a bio page that lists their actual background and links to verifiable external work. I track both for 60 to 90 days and compare Perplexity citation rates.
The gap consistently lands somewhere between 3x and 5x in favour of the named author cluster. The most recent version I ran showed the named-attribution cluster receiving citations 4.1 times more frequently over a 90-day window.
The content had been through the same editorial process. Same word count, same structure, same schema. The only substantive difference was whether a credentialled person was standing behind it. For B2B content especially, that attribution question is not a detail — it is a citation multiplier.
Beyond your own site, B2B brands should be pushing for mentions in trade publications, analyst reports, and professional association resources. These are exactly the types of external references that LLMs draw on when evaluating whether a B2B source is credible.
B2B thought leadership as an AEO asset
Named frameworks and original methodologies are one of the most underused tools in AEO strategy. When you publish something with a specific name — "The Four Stages of Supply Chain Digitisation," say, or your own onboarding methodology — other writers cite it, LLMs encounter it across multiple sources, and your brand starts to get associated with a specific idea rather than just a category. That is entity-building at its most effective.
0.66–0.71
In Ahrefs' cross-platform study of 75,000 brands, branded web mentions showed a strong correlation with AI visibility across ChatGPT, AI Mode and AI Overviews. Named frameworks can help if other people genuinely reference them — the value is the independent discussion, not inventing a branded acronym for its own sake.
Source: Ahrefs, “Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews” (December 2025). Correlation does not prove causation.B2B professional platform strategy
LinkedIn articles are worth more effort than most B2B teams give them. Not shared links or short updates — actual article-length pieces that answer professional questions directly. LLMs pull from LinkedIn content, and an active author profile on LinkedIn strengthens the entity recognition of both the individual and the brand. It is an extra citation surface that costs nothing but time.
B2B-specific schema priorities
Use schema only where it accurately describes visible content. Typical B2B priorities are Organization on the company site, Article/BlogPosting on editorial content, BreadcrumbList for hierarchy, ProfilePage/Person data for authors where appropriate, SoftwareApplication for actual software products, and Dataset for genuine datasets. Schema can clarify entities and enable supported search features, but it is not an AI citation shortcut.
7. B2C AEO/GEO Strategy: The Complete Playbook
Consumer AI search works differently. Queries are shorter, volume is higher, and the buying decision can happen fast — sometimes within the same session as the search. People use ChatGPT and Perplexity to find products, compare options, and solve problems. The content they are looking for is different from what a B2B procurement team needs.
High-intent discovery
AI referrals may be small in volume but often arrive after substantial comparison or problem-solving. For B2C pages, optimise the decision surface: plain-language product identity, accurate price/availability, specifications, pros/cons, fit-for-use guidance, returns/shipping information and trustworthy reviews.
Measure this with your own conversion and assisted-conversion data rather than relying on a universal AI-traffic multiplier.B2C content priorities for AI citation
Content types that drive B2C AI citations
Product category explainers ("what to look for when buying X") · Best-of and top-10 comparison guides · Problem-solution content ("how to fix / treat / do X") · Cost and pricing guides ("how much does X cost?") · Ingredient / material / specification explainers · How-to tutorials tied to product use · Location-specific guides (for local B2C businesses) · Consumer FAQ pages with detailed, specific answers
Queries B2C customers use in AI search
"Best [product category] for [specific need]" · "Is [product/ingredient] safe for [skin type/age/condition]?" · "How to [achieve outcome] at home" · "What's the difference between [product A] and [product B]?" · "How much does [product/service] cost?" · "[Product] reviews: is it worth it?" · "Where to buy [specific product] near me" · "How long does [process/treatment] take?"
B2C product page optimisation for AI citation
For B2C, it is not just blog content that gets cited — product pages get pulled into recommendation and comparison answers too. That means the product description at the top of the page needs to actually answer "what does this do and who is it for?" — not lead with a marketing tagline. Add a specs table in HTML, a FAQ section for the most common product questions, and Product schema with price, availability, and aggregate rating filled in properly.
I did a product page restructure for a consumer goods client in early 2025 — a brand selling direct-to-consumer in a fairly competitive category. Before the work, their product pages led with marketing language: aspirational headlines, benefits-first copy, nothing resembling a functional description of what the product actually was. Perplexity wasn't citing any of them across the 14 target queries we were tracking.
The restructure started with the opening paragraph on each page. Instead of brand positioning, we opened with a factual description — category, what it contains, how it works, who it's for. Dry and clinical on purpose, the kind of sentence a product reviewer would write rather than a copywriter.
Within about five weeks, Perplexity was extracting those opening sentences for several of the queries we'd been tracking. The full restructure — functional description paragraph, HTML specs table, a six-question FAQ, and accurate Product markup — got us to citations on 8 of the 14 target queries within two months. The product hadn't changed. The positioning hadn't changed. The information architecture had. AI systems don't extract marketing language. They extract facts stated once, clearly.
B2C review and user-generated content strategy
AI platforms — especially Perplexity — regularly pull in reviews from Reddit, Trustpilot, and similar platforms when answering B2C product questions. Build your presence on the review platforms your customers actually use. If your site meets Google's review and product structured-data policies, mark up ratings and reviews that are genuinely visible on the page. Never create or aggregate ratings purely to manufacture a rich-result signal.
71%
of Perplexity AI product recommendation answers include at least one citation from a user review platform (Reddit, Trustpilot, or Amazon Reviews), according to a manual analysis of 400 consumer product queries conducted by the IndexCraft research team in November 2025.
Source: IndexCraft Internal Research, "Perplexity Citation Sources for Consumer Product Queries", November 2025 (n=400 queries, available on request)B2C local SEO and AI search
If you have a physical location or serve a specific area, local AI search is worth taking seriously now rather than later. Google AI Overviews are increasingly pulling in local business information for location-intent queries. Get your Google Business Profile properly filled in and kept current. Make sure your NAP (Name, Address, Phone) information is consistent across directories. Add LocalBusiness schema to your contact page on day one.
B2C social proof and community presence
Perplexity regularly surfaces Reddit threads and forum content when answering B2C queries — especially via its Reddit Focus Mode. Being genuinely present in the communities where your customers ask questions (relevant subreddits, Facebook groups, Quora topics, niche forums) does double duty: it builds real relationships with potential customers and puts your brand in the content pool that AI systems draw from.
8. Schema Markup Playbook for New Websites
Schema markup helps search engines classify a page and understand entities, and it can make pages eligible for specific rich-result features. It is useful — but it is not a special AEO/GEO layer. Google explicitly says there is no special schema.org markup required for AI Overviews or AI Mode, and structured data should always match what users can see on the page.
No special “AI schema” required
Prioritise accurate, supported markup over schema volume. Google's current Search gallery supports types such as Article, Breadcrumb, Dataset, LocalBusiness, Organization, Product, ProfilePage and SoftwareApplication. Do not add FAQPage or HowTo everywhere expecting an AI ranking lift; Google no longer lists FAQ as a supported Search rich-result feature, and there is no evidence that FAQ schema is an AI Overview shortcut.
Sources: Google Search Central, structured data documentation and the generative AI optimisation guideImplement schema in this order on a new site:
| Schema Type | Where to Implement | Priority | Notes |
|---|---|---|---|
| Organization | Homepage / organisation page | Day 1 | Use real name, logo, URL and relevant sameAs links |
| Article / BlogPosting | Editorial articles | Day 1 | Author, headline, image, publish/modified dates must match the page |
| BreadcrumbList | Pages with a visible hierarchy | Day 1 | Keep breadcrumb URLs canonical and crawlable |
| ProfilePage / Person | Author profile pages | Month 1 | Use only verifiable credentials and relationships |
| Product | Eligible product pages | Month 1 for ecommerce | Price, availability, shipping/review data must follow policy |
| LocalBusiness | Location / contact pages | Month 1 for local | Keep NAP information consistent with real-world details |
| Dataset | Original datasets and research | When applicable | Describe a genuine dataset; do not use it for ordinary articles |
| SoftwareApplication | Actual software product pages | When applicable | Use properties that match the visible product |
9. Building E-E-A-T from Zero: Practical Steps
E-E-A-T is not a public numeric score you can optimise directly. Treat it as a quality framework: make experience, expertise, reputation and trust easy for people to verify. Named authors, transparent sourcing, accurate dates, first-hand evidence and independent mentions all help users judge whether the page deserves confidence. The full breakdown lives in the E-E-A-T guide.
Experience: how to demonstrate it from day one
Experience signals come from content that could only exist because someone actually did the thing. Screenshots of tools you use. First-person sections describing a specific situation you dealt with. Real numbers, not hypotheticals. Decisions documented with the reasoning behind them. These are the elements that AI-generated content cannot fake — and that AI retrieval systems have learned to value.
The pattern I've seen most consistently with content that gets cited repeatedly is a particular kind of specificity — not general lessons, but accounts of things that actually happened, with real consequences attached. The kind of detail that's hard to fabricate convincingly because it's too particular.
I've been advising B2B clients to include what I loosely call a "what went wrong" section in any content where it's credible — a frank account of a decision that cost something real: time, money, a client relationship. Not "we faced challenges" but "we spent four months and most of a project budget on an implementation that had to be unwound."
One client in an enterprise software vertical wrote honestly about a vendor selection decision that had set them back significantly. That single section was cited in Perplexity responses 11 times in the first 60 days after publication, and earned several editorial mentions from industry publications in their space. Readers trust it because it reads like something that actually happened. AI systems surface it for exactly the same reason: first-person, verifiable, specific.
Expertise: signalling it structurally
Expertise shows up structurally: full author bio pages with credentials and external links; Person schema with jobTitle, worksFor, alumniOf, and knowsAbout; links from author pages to certifications or conference talks. And then there is the content itself — writing at a technical depth that matches the expertise being claimed. A page that positions itself as expert-level but reads like a beginner overview is a red flag for quality raters.
Authoritativeness: building it externally
Authoritativeness is built almost entirely outside your own site — through journalist mentions, guest posts, directory listings, and original research that others cite. A credible external mention can add a kind of independent corroboration that self-published volume cannot create. In months two and three, it is worth slowing down internal content production slightly to spend time on external outreach.
73%
of new websites that earned at least one editorial mention in a DA 60+ publication within their first 90 days achieved their first Google AI Overview citation within 6 months — compared to 24% for sites with no external editorial mentions in the same period.
This data comes from IndexCraft's longitudinal tracking of 89 new-site launches across B2B, B2C, and e-commerce verticals from January to December 2025, with each site's AI citation status checked monthly across Google AI Overviews, Perplexity, and ChatGPT Search.
Source: IndexCraft Internal Research, "External Authority Signals and AI Overview Citation Rate", 89 new-site launches, January–December 2025 (methodology available on request)Trustworthiness: the non-negotiables
Trustworthiness comes down to a handful of non-negotiables: HTTPS, a clear About page, an editorial policy, named authors on every piece, inline citations linked to primary sources, accurate publication dates that get updated when content changes, and a privacy policy. If you are writing about health, finance, or legal topics, add a disclaimer and a named expert reviewer. These are not extras — they are what separates a trustworthy site from one that quality raters flag.
10. Link Building for AI Citation Authority
Links still matter for discovery, page-level authority and referral traffic, but AI visibility is not well explained by link volume alone. Recent Ahrefs research found off-site brand mentions correlated more strongly with AI brand visibility than backlink count or Domain Rating. For a new site, think “earned reputation” rather than “link quota.”
Mentions > link volume
Ahrefs found branded web mentions correlated strongly with AI visibility (roughly 0.66–0.71 across major platforms), while classic link metrics were weaker predictors. That does not make backlinks irrelevant; it means the goal should be useful third-party coverage and citations, not bulk link acquisition.
Source: Ahrefs, “Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews” (December 2025)Original data and research
Even a small, well-designed survey can generate citable statistics if the sample, recruitment method, questions and limitations are disclosed. When you are the primary source of a number that other writers want to use, you earn editorial links naturally. Put the findings on a dedicated page, tell journalists and bloggers in your space about it, and reference your own data in other articles so people encounter it organically.
The highest single-action return I've seen on a new site in its first year was a small original research piece — not a lengthy industry report, just a focused survey of around 55 people in a specific professional role, asking questions that practitioners in that space would actually find useful to have answered.
I've helped a few clients do this in the first three months of a new site. The approach is the same each time: identify a question your audience genuinely debates or is uncertain about, build a short survey of 10 to 12 questions that takes under five minutes to complete, find respondents in relevant professional communities, and publish the findings with a clear headline stat.
The finding needs to be genuinely surprising — not "people care about X" but "the thing practitioners say is the real problem is Y, not what most people in the space assume." When you have that, it travels.
The most recent version I ran with a client generated 11 editorial mentions within eight weeks of publication, and a Perplexity citation for a statistics query in their space within the first month. For a site with no domain authority, nothing else I've tried replicates that return in the same timeframe.
Expert roundups as a link acquisition vehicle
Expert roundup articles — collecting quotes from ten to fifteen industry names on a specific question — work well for new sites precisely because they give something back to the people quoted. Experts typically share and link to pieces they are featured in. Use LinkedIn to find and contact them. Personalise every message. Generic outreach gets ignored.
Resource page link building
Many established sites keep curated "resources" or "useful links" pages. Find them with searches like "resources" + [your topic] or "useful links" + [your topic]. Email the owner with a specific reason why your page adds value to their list. Response rates are usually low and vary a lot by niche, but the links you earn this way tend to be high-quality and stick around.
Journalist request services
HARO (Help a Reporter Out, relaunched in April 2025 under Featured.com after its 2024 shutdown) and Qwoted connect journalists seeking expert sources with subject-matter experts. Responding promptly and substantively to relevant journalist requests earns editorial mentions and links from news sites and industry publications — exactly the type of authoritative external reference that strengthens AI search citation authority.
11. The Highest-ROI Content Formats for New Sites
New sites need content that does two things at once: build organic rankings (the path to Google AI Overview citations) and get picked up directly by real-time crawlers like Perplexity. In my experience these formats give a new site the best return on both goals. Treat the order as practitioner judgment, not a benchmark.
Useful when a term genuinely needs explanation. Lead with a plain-language definition, then add context, examples, edge cases and first-hand insight. There is no universal optimal word count and no need to add FAQ content to every sub-question.
HTML tables make AI extraction simple. Include a clear "verdict" or "best for" section. Update quarterly as products/prices change. Both B2B and B2C perform well here. Often picked up by Perplexity.
Earns natural editorial backlinks. Build as a "living" page — refresh annually. Name and hyperlink every statistic's source. Use Dataset schema where applicable. Relatively low competition for new sites.
FAQ sections are useful when readers genuinely have follow-up questions. Keep answers self-contained and specific, but do not copy People Also Ask mechanically or publish FAQs just to add schema. Google no longer lists FAQ as a supported Search rich-result feature.
Earns external links and brand mentions. Even 50-person surveys add significant value. Promotes brand entity recognition in LLMs. Announce via press release for earned media. Refresh annually for continued relevance.
Competition is often lower here, which makes it a sensible entry point. Each term is an independent citation target. Internal link hub for the entire content cluster. Drives organic traffic from definitional queries. Scales well: add terms incrementally over time.
12. Perplexity: A Useful Early Discovery Test for New Sites
Perplexity is useful for testing whether a new page is being discovered and retrieved for a narrow question, but there is no guaranteed “two-to-four-week” window. Treat it as one visibility surface among several. A new page can sometimes be cited before it has strong Google rankings, especially when it is highly specific and the existing source set is weak, but results vary by topic and query.
780M+ queries a month
Perplexity's CEO said the platform handled about 780 million queries in May 2025, up from roughly 230 million in August 2024. Third-party trackers now estimate more than a billion a month in mid-2026. Perplexity does not publish audited usage numbers, and user estimates vary widely: from the mid-30 millions of monthly active users on the core product to 100 million or more across all its products.
For new sites, the practical value of Perplexity is that citation testing is easy to observe directly. Do not mistake fast feedback for a universal ranking advantage.
Sources: Perplexity CEO figures as reported in the press; third-party trackers (Gradually, getAIrefs), mid-2026. Estimates differ by source.Three sensible checks for Perplexity visibility are: keep its search crawler unblocked if you want inclusion, answer the target question directly, and publish information that is meaningfully more useful or current than what is already available. Freshness matters when the query itself is freshness-sensitive; it is not a substitute for accuracy or authority.
This is something I started doing systematically after noticing the same pattern on enough articles to feel confident it wasn't a coincidence.
For any article I want to appear in Perplexity, I open Perplexity and ask my target question directly. If it uses a competitor's content instead of mine, I read the specific sentence Perplexity extracted from their page and look at its structure.
Across about 25 articles where I've done this check, the extracted sentence almost always follows the same pattern: it starts with the answer noun, includes at least one number or specific named entity, and runs somewhere between 20 and 32 words.
The pattern I look for is simple: useful information should arrive early. Cut empty scene-setting such as “in today's landscape”, but keep the nuance a human needs. A direct first sentence plus a short caveat often reads better — and is easier to quote — than either marketing fluff or robotic one-line answers. That extractable sentence structure is now the first thing I look at when reviewing any article opening paragraph.
13. Tools to Track Your AEO/GEO Progress
GEO tracking is still catching up to traditional SEO measurement, but you can build a useful stack from mostly free tools in your first year.
| Tool | What It Tracks | Cost | Priority |
|---|---|---|---|
| Google Search Console | Indexation status, organic rankings, AI Overview impressions and click data | Free | Essential |
| Bing Webmaster Tools | Bing indexation, crawl health, rankings and AI Performance citations / grounding-query insights | Free | Essential |
| Google Analytics 4 | Referral traffic from AI platforms: perplexity.ai, chat.openai.com, gemini.google.com | Free | Essential |
| Manual AI Query Testing | Direct citation checking — query each platform manually for your 20 target queries monthly | Free (use free tiers) | Essential |
| Schema Markup Validator | Validates JSON-LD schema syntax before deployment; use before every schema push | Free (validator.schema.org) | Essential |
| Ahrefs / Semrush | Backlink tracking, keyword rankings, competitor content gap analysis | Paid ($99–$250/mo) | Recommended Month 4+ |
| Brand24 / Mention | External brand mentions across web, news, and social — essential for entity tracking | Paid ($49–$99/mo) | Recommended Month 3+ |
| AnswerThePublic | Question mapping for content brief development; visual query clustering | Free tier available | Recommended |
14. Critical Mistakes New Sites Make with AEO/GEO
These are the mistakes I see repeatedly across new sites trying to build AI search visibility. Each one is expensive in time, wasted content budget, or delayed authority — sometimes all three. They come from direct observation across 47 new-site audits between January and December 2025.
❌ Mistake #1: Publishing thin content at high volume
- Publishing dozens of low-value pages just to cover every keyword or prompt variation. Google explicitly warns that scaled pages created mainly to manipulate rankings or generative AI responses can violate spam policy.
- There is no magic 600-, 1,500- or 2,500-word threshold. Use the length needed to solve the task, then remove repetition. Add first-hand evidence, examples, visuals or data when they genuinely improve the answer.
❌ Mistake #2: Ignoring Bing for ChatGPT visibility
- Do not optimise only for one discovery system. Keep Google and Bing indexing healthy, and manage AI-specific crawler access separately. For ChatGPT Search specifically, OpenAI documents OAI-SearchBot as the crawler to allow for summaries and snippets; do not substitute generic “Bing powers ChatGPT” advice for OpenAI's own crawler guidance.
- In a manual crawl test I ran in October 2025 across 34 recently-launched sites that had excellent Google indexation, 12 of them (35%) had not submitted their sitemap to Bing Webmaster Tools — and all 12 had zero ChatGPT Search citations across their target queries. The fix took under 10 minutes per site. It was a small, informal sample and shows correlation only, so read it as a reason to set up Bing Webmaster Tools, not as proof that a Bing submission causes ChatGPT citations.
❌ Mistake #3: No named authors on any content
- Anonymous content cannot establish E-E-A-T. If your articles are bylined to "Staff Writer" or "Admin" with no linked bio page, AI systems cannot verify any expertise or authority behind the content.
- Google's Search Quality Rater Guidelines explicitly flag anonymous authorship as a negative quality signal for YMYL (Your Money Your Life) and advice content. Every piece of content should have a named, credentialled author with a linked bio page.
❌ Mistake #4: Building content without a topic cluster architecture
- Publishing articles on a wide variety of loosely-related topics — hoping to catch traffic wherever possible — delays the development of topical authority that AI search systems require.
- Based on IndexCraft's analysis of 89 new-site launches, sites with defined topic cluster architecture achieved their first AI Overview citation an average of 4.2 months faster than sites with unstructured content publishing strategies. This is an internal observation, not an industry benchmark.
❌ Mistake #5: Using AI-generated content without adding genuine original value
- Publishing content that was entirely AI-generated without any original examples, data, insights, or editorial perspective produces content that lacks the specific signals (firsthand experience, original data, named credentials) that AI search systems use to select sources worth citing.
- Google's own guidance is that the tool used to draft a page is not the problem; scaled, low-value content is. Use AI as a drafting aid, and make sure the published page carries real expertise, examples and accountable authorship.
15. Priority Action Matrix by Business Type and Timeline
Use this to work out where to focus based on your business type and where you are in your timeline. Do the High items first. Medium and Low can wait.
| Action | B2B New Site | B2C New Site | Timeline |
|---|---|---|---|
| Google Search Console + sitemap submission | High | High | Day 1 |
| Organization schema on homepage with sameAs | High | High | Day 1 |
| Named author bio pages with Person schema | High | Medium | Day 1 |
| First topic cluster (pillar + 4–6 supporting pages) | High | High | Month 1 |
| Useful FAQ sections where intent warrants them | Medium | Medium | As needed |
| Product schema + AggregateRating on product pages | Low | High | Month 1 |
| LocalBusiness schema + Google Business Profile | Low | High (local B2C) | Month 1 |
| Bing Webmaster Tools setup + sitemap submission | High | High | Month 1 |
| Statistics roundup page with sourced citations | High | Medium | Month 2 |
| Industry directory listings (minimum 3) | High | Medium | Month 2–3 |
| Original research publication (survey / study) | High | Medium | Month 3 |
| HARO / Qwoted journalist outreach | High | Medium | Month 2 onwards |
| Professional platform article publishing by named authors | High | Low | Month 2 onwards |
| Review platform presence (G2/Capterra or Trustpilot) | Medium | High | Month 2 onwards |
| Second topic cluster (pillar + 4–6 supporting pages) | High | High | Month 4–5 |
| Monthly AI citation audit (manual, 20 queries) | High | High | Month 1 onwards |
| Thought leadership framework / named methodology | High | Low | Month 3–6 |
Conclusion: Building AI Search Authority Is a Compounding Investment
The sites that will be cited consistently in AI search two years from now are mostly already building their foundations. Every genuinely useful page gives you another relevant entry point. Every credible external mention strengthens your reputation. Accurate structured data can clarify the page, and repeat visibility gives you more evidence about what users and retrieval systems find useful. The compounding effect comes from a better body of work and stronger reputation — not from collecting technical “GEO points.”
New websites are behind, but not permanently. That gap closes faster for sites that get the architecture right from day one than for established sites trying to retrofit their legacy content strategy around AI search. You have one advantage they do not: you can build this correctly from the start.
This shift is still moving quickly. That is exactly why the safest strategy is to build on things that survive platform changes: useful content, strong technical SEO, first-hand expertise, measurable user value and a reputation that exists beyond your own domain.
Frequently Asked Questions
Can a brand new website appear in AI search?
Yes, but there is no guaranteed timeline. For Google AI Overviews and AI Mode, the page first needs to be indexed and eligible to appear in Google Search with a snippet. Other products use their own retrieval systems. A narrow, genuinely useful new page can sometimes be discovered early, but reputation, competition and query type all affect visibility.
What is the difference between AEO and GEO?
AEO is usually used for making content easy for answer systems to understand and quote; GEO focuses more specifically on visibility and citation inside generative AI experiences. In practice, the foundations overlap heavily with good SEO: crawlability, helpful content, clear structure, evidence and reputation. Google explicitly treats optimisation for its generative Search features as part of SEO.
How long does it take a new website to appear in Google AI Overviews?
There is no dependable “three-month” or “six-month” rule. Google does not publish a waiting period. The page must be indexed and eligible for Search, and then it still competes on relevance and quality. Track actual impressions in Search Console's Generative AI performance report instead of planning around an arbitrary age threshold.
Is AEO/GEO strategy different for B2B and B2C?
The foundations are the same, but the evidence people need is different. B2B content usually benefits from technical depth, implementation constraints, case evidence and expert authorship. B2C content often needs clear product facts, comparisons, price/availability, reviews, local information and visual proof. Start from the decision the user is trying to make.
Do I need schema markup on every page?
No. Add structured data when it accurately describes the visible page and is useful for a supported search feature or entity definition. Typical priorities are Organization, Article/BlogPosting, BreadcrumbList, Product, LocalBusiness, ProfilePage and Dataset where applicable. There is no special “AI schema” required for Google AI Overviews or AI Mode.
What should a brand new site publish first?
Publish the smallest set of pages that answers the core questions a real buyer or reader needs before taking the next step. Usually that means one strong hub page plus a few distinct supporting guides, comparisons or tutorials. Avoid launching dozens of thin keyword variations. Add something competitors cannot easily copy: first-hand examples, original data, templates, screenshots or expert judgment.
Does AI-generated content hurt SEO, AEO or GEO?
Using AI tools is not automatically a problem. Google says the issue is scaled or low-value content, not the tool used to draft it. AI can help with research and structure, but the published page should add real value — accurate sourcing, original experience, editorial judgment, useful examples and a reason for someone to choose it over a generic summary.
Sources and further reading
| Source | Why it matters |
|---|---|
| Google Search Central — Optimizing your website for generative AI features | Official 2026 guidance: SEO fundamentals remain relevant; query fan-out is used; unique, non-commodity, people-first content matters; no special AI markup or llms.txt is required for Google Search. |
| Google Search Central — AI features and your website | Pages need normal Search eligibility; there are no additional technical requirements for AI Overviews or AI Mode. |
| Google Search Central — Generative AI performance reports in Search Console | Dedicated AI visibility reporting for Search and Discover; rolled out to all websites worldwide by August 31, 2026. |
| OpenAI — Publishers and Developers FAQ | OAI-SearchBot controls ChatGPT Search discovery for summaries/snippets; GPTBot is a separate training crawler. |
| Bing Webmaster Blog — AI Performance | Bing Webmaster Tools reports AI citations, cited pages and citation activity across supported Microsoft AI experiences. |
| Bing — AI Visibility Insights | Adds intents, topics, citation share and comparison views to help diagnose why content is cited. |
| Ahrefs — 38% of AI Overview Citations Pull From The Top 10 | 37.9% of cited URLs also appeared in the first 10 SERP blocks in a March 2026 study of 863K SERPs and 4M AI Overview URLs. |
| Ahrefs — Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews | Branded web mentions correlated strongly with AI visibility (0.66–0.71). The authors explicitly caution that correlation is not causation. |
| Google Search Central — Structured data introduction | Structured data helps Google classify content and can enable supported rich results; JSON-LD is recommended when practical. |
| Google Search Central — Guidance on generative AI content | AI-assisted content is not inherently disallowed; scaled content without added value can violate spam policy. |
| IndexCraft first-party tracking | Where this guide uses IndexCraft observations (for example, launch-cohort citation timing), they are labelled as internal and directional. Publish the methodology publicly before presenting these observations as industry benchmarks. |