How to Optimize Content for AI Search: What the Evidence Supports

How to optimize content for AI search, in one line: make sure the crawlers behind ChatGPT, Perplexity, Claude and Google can reach your pages, put the content in the HTML they read, and state your key facts plainly and identically everywhere they appear. No formatting trick (schema, FAQ blocks, "chunking", llms.txt) has evidence of getting a page found more often, and the controlled tests of AI-targeted edits found most do nothing or do harm.

That is a smaller promise than most guides make. It is also the one the evidence supports, and every step below is something you can check today.

How AI search finds and uses your content

Every AI answer that cites a web page passes three gates. Your content can only influence the gates it reaches.

1. Access. A crawler must be allowed to fetch the page, and your firewall must let it through. Fail here and nothing else on this list matters.
2. Retrieval. For each question, the AI system runs its own searches and picks candidate pages. OpenAI says that when ChatGPT search uses a partner search provider, it "typically rewrites your query into one or more targeted queries," and Google says AI Overviews and AI Mode may use "query fan-out." Topical relevance is the one factor the research agrees on, though it was measured on pages that had already been retrieved.
3. Citation. Among the pages retrieved, the model decides which to use and which to cite. Nearly every published "GEO" experiment happens here.

A fourth source skips all three gates: what the model memorized in training, frozen at a knowledge cutoff. A July 2026 critical survey of GEO research (a single-author arXiv preprint) concludes: "already-retrieved content can causally alter its citation or use, but no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability." In plain terms, nobody has proven a content tactic that gets you found more often.

Step 1: Let the AI crawlers in

Each AI company runs separate crawlers for separate jobs, and blocking the wrong one removes you from a product. The ones that matter for AI search:

Sources: OpenAI's crawler documentation, Perplexity's crawler documentation, Anthropic's crawler help page, Google's crawler list and Meta's crawler documentation.

Your robots.txt file (at yourdomain.com/robots.txt) tells each crawler what it may fetch. A setup that stays in AI search but opts out of OpenAI training looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: Meta-WebIndexer
Allow: /

User-agent: GPTBot
Disallow: /

robots.txt is only half the gate

OpenAI's ChatGPT search help makes eligibility a two-part test: "allow OAI-Searchbot to crawl the site and confirm that the website host or content delivery network allows traffic from OpenAI's published searchbot IP addresses." Firewalls and bot protection can turn crawlers away with a 403 error while robots.txt says yes. Our post on the Cloudflare AI crawler default covers one common source of that block.

A quick test from your own computer: request a page while sending the user agent OpenAI publishes for its search crawler, and read the status code the command prints. A 403, 429 or 503 means something in front of your site is blocking that user agent.

curl -s -o /dev/null -w "%{http_code}" -A "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36; compatible; OAI-SearchBot/1.4; +https://openai.com/searchbot" https://www.example.com/

On Windows, type curl.exe and replace /dev/null with NUL. That string is OpenAI's published example and its version number changes, so copy the current one from developers.openai.com/api/docs/bots. A 200 from your own computer is not proof, because your firewall may treat the real crawler's IP addresses differently; a 403 is a strong warning.

Check you have not hidden your own content

Google's AI features documentation names the controls that limit what its AI shows from a page: nosnippet, data-nosnippet, max-snippet and noindex. These often arrive by accident, inherited from a template. Search your page source for them.

Step 2: Put the content in the HTML

Many AI crawlers read the raw HTML your server sends and do not run JavaScript. Vercel's analysis of crawler traffic found that ChatGPT and Claude crawlers fetch JavaScript files but "don't execute them. They can't read client-side rendered content" (Vercel, December 2024).

Googlebot and AppleBot do render pages, so the risk sits with crawlers like OpenAI's and Anthropic's. To check a page, open it in your browser and press Ctrl+U (Cmd+Option+U on a Mac) to view the source. Search for a sentence or a price from the page.

If it is not in the source, it was added by JavaScript, and a crawler that does not render may never see it. Ask your developer to render that content on the server.

Step 3: Answer the question the page is about, in plain sentences

Google is explicit that you "don't need to write in a specific way" for generative AI search (Google's AI optimization guide, July 2026). The survey above found the most reproducible factors were "topical relevance and context position," and you cannot control context position. So write each page about one thing, and say the thing.

The practical test is whether a sentence could be checked. A model can only repeat a fact you actually state; vague copy leaves the gap for other sources to fill.

This is not a citation trick. It is the minimum for being described correctly once you are retrieved.

Step 4: Make every fact match everywhere you publish it

If your pricing page says $49, your FAQ says $59 and a 2024 blog post says $39, an AI system that reads all three has no way to know which is current. Contradictions on your own site are the most fixable cause of wrong AI answers, because you control every copy.

Search your domain for each fact's old values, for example site:example.com "$39" in Google. Fix every copy in one release, including structured data and footer text. Our guide to brand entity optimization covers the full fact list, and how often to update content for AI covers when to review it.

Step 5: Be described correctly where AI already looks

AI search leans on a small set of heavily cited news outlets. A 2025 study of 65,000 AI search responses found that "News citations concentrate heavily among a small number of outlets" (arXiv:2507.05301). That is an association with outlet size, measured on news only, and not a lever you can pull this quarter.

What you can do is make sure the third-party pages AI may retrieve instead of yours say the same thing yours do. Update your listings on review sites, marketplaces, directories, LinkedIn and Crunchbase, and ask partners to correct old descriptions. Corroboration means your facts agree wherever they appear.

What does not work, and what the evidence says

One caveat cuts the other way. Most of these tests used pages that were already being retrieved or cited, so whether any of them helps a page enter the candidate set at all is untested. Untested is not the same as proven, and none of these is worth prioritizing over Steps 1 to 4.

How to optimize content for Google AI Mode and AI Overviews

Google's answer is short: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." A page must be "indexed and eligible to be shown in Google Search with a snippet." So the checklist is classic Search hygiene:

• The page is indexed (check it with the URL Inspection tool in Search Console).
• No noindex, nosnippet or max-snippet:0 on pages you want shown.
• Googlebot is allowed in robots.txt and not blocked by your firewall.
• The page is useful on its own terms. Google's guide says its AI features are "rooted in our core Search ranking and quality systems."

How to check whether it worked

Ask ChatGPT, Perplexity, Gemini, Claude and Google the questions your buyers ask, and record the date, the answer and the cited sources. A cited answer came from a live search; an uncited one may come from memory, which no content change reaches until the next model. OpenAI itself warns that "Search results and citations can be incomplete, outdated, or incorrect."

For a quick read on ChatGPT and Google AI Overviews, run the free AI visibility checker. 99Visibility's full audit also tests robots.txt for 18 AI and search crawlers, requests your page as each one to catch firewall blocks, and scans for snippet suppression.

Frequently asked questions

How do I optimize content for Google AI Mode?

The same way you optimize for Google Search. Google states there are no additional requirements or special optimizations for AI Mode or AI Overviews: the page must be indexed and eligible to show a snippet. Check indexing in Search Console and remove any accidental nosnippet or noindex.

How do I get cited in ChatGPT?

Allow OAI-SearchBot in robots.txt, make sure your firewall lets OpenAI's search crawler through, and keep the facts on your pages plain and consistent. That makes you eligible; OpenAI says only that ChatGPT ranks results "using multiple factors" and that "Placement is not guaranteed." No one can promise a citation.

Can ChatGPT write SEO content?

It can draft it. Google's guidance on generative AI content says AI "can be particularly useful when researching a topic, and to add structure to original content," but warns that mass-producing pages without adding value may violate its scaled content abuse policy. It also says it is "critical to manually factcheck and review all AI-generated content." An AI-drafted page with a wrong price is one more wrong source.

Does schema markup help content get cited by AI?

Not on current evidence. Google says no special markup is needed for its AI features, and a matched test of 1,885 pages that AI Overviews already cited (Ahrefs) found no major citation uplift on any platform. Use schema for Google rich results, and keep it identical to the visible page.

Is writing for AI search different from writing for SEO?

No. Google says you do not need to write in a specific way for generative AI search, and the controlled studies found AI-targeted rewrites ineffective or harmful. Write for the reader, state facts plainly, and spend the effort on crawler access and fact consistency instead.

See whether AI crawlers can reach your site and what AI says about your brand with a free 99Visibility audit.