What Elements Are Foundational for SEO With AI Search

Each foundation below has a mechanism that the engine operators document themselves. Most work by removing a block, not by adding a signal. You can't make a model choose you, but you can stop being ineligible.

A critical survey of 45 GEO studies (arXiv 2607.14035, July 2026) found no technique with a stable, cross-platform causal effect on discoverability. The access controls above sit outside that problem. Their consequences are stated by OpenAI, Google, Bing and others in their own documentation.

An AI answer passes through four stages. Knowing which one you're working on stops you wasting effort.

1. Parametric. What the model knows with no web access, frozen at a training cutoff. OpenAI lists a February 16, 2026 cutoff for its GPT-5.6 models and August 31, 2025 for the Instant path. Nothing you publish this quarter reaches these answers.
2. Access. Whether a crawler may fetch and store your page. This is where the foundations live, and you control it.
3. Retrieval. Whether your page enters the model's context for a given query. Retrieval fires per query, at the engine's discretion.
4. Generation. Which retrieved pages get used and cited. The model decides.

Strong classic SEO helps but does not settle stage 3. Ahrefs found that 37.9% of URLs cited in AI Overviews also appeared in the top 10 results (March 2026). That's why strong SEO alone doesn't guarantee AI visibility. Fail stage 2, though, and nothing later matters.

HTML-rendered content

Vercel's crawler study found that ChatGPT and Claude crawlers fetch JavaScript files (11.50% and 23.84% of requests) but don't execute them. Independent testing in June 2026 confirmed the same for OAI-SearchBot, ClaudeBot, PerplexityBot and others. Googlebot renders JavaScript, so this gap mostly affects non-Google engines. If your pricing only appears after hydration, those crawlers never see it. Server-side rendering beats schema as a lever for exactly this reason.

Crawler permission

robots.txt rules are per bot. OpenAI states that each setting "is independent of the others". GPTBot governs training use. OAI-SearchBot governs ChatGPT search, and sites that block it "will not be shown in ChatGPT search answers". You can allow search while declining training:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

Check PerplexityBot, Claude-SearchBot and Meta-WebIndexer the same way. Meta says allowing Meta-WebIndexer "helps us cite and link to your content". There is no robots.txt token for Google AI Overviews or AI Mode: Googlebot rules cover them.

Network reachability

OpenAI's help centre says eligibility needs two things: allow OAI-SearchBot, and confirm your host or CDN allows its published IP addresses. A firewall can block a crawler that robots.txt allows, often with a 403 nobody sees. Since September 15, 2026, Cloudflare blocks Training and Agent crawlers by default on ad-displaying pages for new domains and existing free-tier customers. A site can lose access without anyone editing it.

Indexability

Google says a page must be "indexed and eligible to be shown in Google Search with a snippet" to appear as a supporting link in AI Overviews or AI Mode, with "no additional technical requirements". Bing's index backs Microsoft Copilot. Check each index separately, because one being healthy tells you nothing about the other.

Also check Search Console's generative AI control where it's available. Inclusion is the default, but a property can be opted out, and child properties inherit the parent's setting.

Snippet eligibility

noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers on Google. Bing says data-nosnippet content stays indexed but is excluded from snippets and AI summaries. These controls are useful when deliberate. They often arrive by accident, inherited from a template.

Consistent facts

A specific fact (a price, a number, a one-line definition) on a fetchable page gives an engine a sentence it can quote. That is the only positive experimental signal for on-page content in the evidence. When your pricing page and footer disagree, you've handed the engine two answers. Fix the contradiction at the source, and see how to manage your brand as an entity for facts that live on other sites.

Both sound foundational. Neither is something an AI engine reads.

• E-E-A-T. Google says E-E-A-T "isn't a specific ranking factor". It is a concept that quality raters use, and their ratings do not position any page. There is no score to improve. The habits behind it, such as naming authors and citing sources, can still be sound practice. See Google's helpful content guidance and E-E-A-T.
• Domain authority. Domain Authority is a Moz metric, and similar scores come from other SEO vendors. No engine computes them, and Google publishes no site-level authority score. Brand prominence is strongly associated with citation, but these metrics are proxies for brand size, not a lever.

The same logic applies to schema and llms.txt. Google says structured data isn't required for generative AI search. Ahrefs' matched test found AI Overviews citations fell 4.6% after pages added schema (those pages were already heavily cited). Google Search ignores llms.txt. Don't rip out existing markup, since Google says it does no harm. Just don't count it as a foundation.

In your control: rendering, robots.txt rules, firewall and CDN settings, index and snippet directives, and the facts on your own pages.

Partly in your control: being named on the third-party sites that engines cite instead of you. The evidence shows an association there, with no promised effect.

Not in your control:

• What the model learned before its cutoff.
• The queries the engine actually runs. OpenAI says ChatGPT search rewrites your prompt into one or more targeted queries, and Google may fan out across subtopics.
• Whether retrieval fires for a given answer.
• Where your page lands in the model's context, and whether it gets cited.

So set expectations plainly. Fixing the foundations removes reasons you can't appear. It does not promise a citation, a ranking or traffic.

Work from the gate that blocks most to the one that blocks least.

1. Rendering. Fetch your key pages with JavaScript off. Confirm that prices, product names and headings are in the raw HTML.
2. robots.txt. List the rule each AI search bot gets. Separate training bots (GPTBot, ClaudeBot, Google-Extended) from search bots (OAI-SearchBot, PerplexityBot, Claude-SearchBot).
3. Reachability. Request pages as each crawler and look for 403s and challenge pages. The AI Crawler Checker does this per crawler.
4. Index status. Check Google Search Console and Bing Webmaster Tools separately, including the generative AI control.
5. Snippet controls. Find every nosnippet, max-snippet and data-nosnippet and confirm each one is intentional.
6. Facts. List your ten most important facts and make each one read identically on every page.
7. Measure per engine. The free checker on this site shows a snapshot of whether AI engines name you for three buyer questions.

For the full list, use the AI SEO checklist. Recheck access after any CDN, hosting or template change.

• A blanket AI block. Blocking every AI user agent to stop training also removes you from ChatGPT search.
• Allowing GPTBot to get found. GPTBot governs training only. Search uses OAI-SearchBot.
• Trusting robots.txt alone. An allow rule means nothing if the firewall returns 403.
• Treating schema, llms.txt or chunking as foundations. Google's generative AI optimization guide says none of them is needed.
• Mass-publishing thin pages. In Otterly's 2026 experiment, Google deindexed two sites of about 1,000 AI-generated posts each, with no manual action reported.
• Reading one blended score. In the same test, Copilot cited a site thousands of times while ChatGPT barely did. Measure each engine separately.
• Hidden text or prompt injection aimed at AI crawlers. It is manipulation, it violates engine policies, and it puts your index status at risk. Never use it.