AI SEO: How to Get Found and Cited in AI Search

There is no single AI SEO switch. Google says nothing extra is required beyond being indexed and eligible for a snippet. ChatGPT, Perplexity, Gemini and Copilot each have their own controls, so you decide bot by bot. A realistic first month is four jobs: let the right bots in, fix what stops pages being read, rewrite key pages so they answer questions directly, and measure.

What is AI SEO?

It sits on top of classic search engine optimization (SEO). You still need pages that can be crawled and indexed, and content that answers real questions. What changes is who reads your page first: often a machine that summarizes it for a person.

You will also see this called AI search optimization, answer engine optimization (AEO) or generative engine optimization (GEO). GEO is the research-led label for improving how generative AI systems use and cite your content; the theory and evidence are in our Generative Engine Optimization guide. This page is the practical side: what each AI engine says it does, which switches you control, and what to do in your first 30 days.

The terms you need:

• AI Overviews are the AI-written summaries Google shows at the top of some search results.
• AI Mode is Google’s conversational search experience. Alphabet reported in July 2026 that AI Mode had passed 1 billion monthly active users since its global expansion the previous October.
• A crawler (or bot) is a program that visits web pages automatically. AI companies run several kinds: some collect content to train models, some fetch pages to build search answers, and some fetch a page only when a person asks.
• robots.txt is a plain text file at the root of your site that tells crawlers which parts they may visit.
• A citation is a link or source label an AI answer attaches to a statement.

Is it worth the effort?

The audience is real, and it is growing alongside search, not instead of it.

The Reuters Institute describes the growth of chatbot use for news as “fast rather than explosive”. People are adding AI to search, not dropping search.

How does each AI search engine find and use your content?

Each AI search engine has its own crawlers and its own rules, so there is no single AI SEO switch. The table summarizes what the vendors’ own documentation says, as we read it on 2026-10-10. Where we found no official guidance, the table says so.

Engine	What it draws on	Crawler and controls (per vendor docs)	What the vendor says
Google AI Overviews and AI Mode	Google Search index	Normal Googlebot rules; snippet controls (nosnippet, data-nosnippet, max-snippet, noindex)	No additional requirements; no special markup needed; page must be indexed and snippet-eligible
ChatGPT search	OpenAI’s search index	OAI-SearchBot for search, GPTBot for training, ChatGPT-User for user-requested fetches	Opted-out sites will not be shown in ChatGPT search answers; robots.txt may not apply to ChatGPT-User
Perplexity	Perplexity’s search index	PerplexityBot for search, Perplexity-User for user actions	PerplexityBot is not used for foundation-model training; Perplexity-User generally ignores robots.txt
Gemini	Google models, with grounding	Google-Extended robots.txt token (no separate user-agent string)	Controls use for training future Gemini models and for grounding; does not affect Search inclusion
Microsoft Copilot	Bing Webmaster Tools reports on content cited in Copilot, Bing AI summaries and select partner integrations	Bing says it respects robots.txt and other supported controls	Microsoft’s AI Performance report shows how often your pages are cited; it advises clear structure, fresh content and IndexNow

Google AI Overviews and AI Mode

Google says there are no extra steps. Its documentation states that “there are no additional requirements to appear in AI Overviews or AI Mode” and that you do not need special schema.org structured data. A page must be indexed and eligible to be shown with a snippet (the short text excerpt Google displays under a result).

Google’s separate guide on optimizing for generative AI features adds more:

• You do not need new machine-readable files, AI text files, markup or Markdown to appear in Google Search.
• There is “no requirement to break your content into tiny pieces”.
• Structured data “isn’t required for generative AI search”.
• llms.txt “will neither harm nor help” Search visibility.
• Seeking inauthentic mentions of your brand across the web “isn’t as helpful as it might seem”.

What you control: the standard snippet tools. nosnippet and data-nosnippet stop text being quoted, max-snippet limits how much, and noindex removes the page from Search. Traffic from AI features is counted in Search Console under the “Web” search type, so it is blended with normal results.

How often AI Overviews appear also swings. Semrush tracked more than 10 million keywords in 2025:

That is vendor data, but it shows why one month’s snapshot should not drive a strategy.

ChatGPT search

OpenAI documents three separate crawlers. The split matters more than anything else on this page:

• GPTBot collects content that may be used for training.
• OAI-SearchBot powers ChatGPT search. OpenAI says opted-out sites “will not be shown in ChatGPT search answers”, and that the change takes about 24 hours to apply.
• ChatGPT-User fetches a page when a person asks ChatGPT to open it. OpenAI says robots.txt rules may not apply to it.

So blocking GPTBot does not remove you from ChatGPT search, because that is a different bot. Blocking OAI-SearchBot does. For a sense of how common the choice is, the HTTP Archive’s 2025 Web Almanac found GPTBot named in the robots.txt of 4.5% of desktop sites (2.9% in 2024). That counts sites that mention the bot, not sites that block it.

On scale, Similarweb’s panel shows ChatGPT’s share of generative AI web visits shrinking while Gemini’s grows:

Perplexity

Perplexity’s documentation lists two bots. PerplexityBot is used to surface and link websites in Perplexity search results and, Perplexity says, is not used to train foundation models. Perplexity-User supports user actions; because those fetches are triggered by a person, Perplexity says it “generally ignores robots.txt rules”.

Whether bots really follow robots.txt is contested. Here is who claims what:

• Documented by the vendor: the Perplexity-User behavior above.
• Alleged by a third party: in August 2025, Cloudflare reported that Perplexity used undeclared crawlers that impersonated a normal Chrome browser, and that tests on new domains with robots.txt set to disallow-all still returned content details. Perplexity publicly disputed it, arguing that assistants acting on a user’s request are not crawlers (as reported by press coverage; we could not open Perplexity’s own post).
• Measured independently: an academic preprint studying 130 self-declared bots found that stricter robots.txt directives are less likely to be obeyed and that some categories, including AI search crawlers, rarely check robots.txt at all.

robots.txt is a polite request, not a lock. If content must stay private, put it behind a login.

Gemini

Gemini is Google’s AI assistant. Alphabet reported 950 million monthly active users for the Gemini app in July 2026. For site owners the control is Google-Extended, a robots.txt token (not a separate crawler name).

Google says it controls whether your content is used to train future Gemini models and for grounding, meaning checking an answer against live content. Google also states that it “does not impact a site’s inclusion in Google Search nor is it used as a ranking signal”.

Microsoft Copilot

Microsoft’s guidance is about measurement and basics. In February 2026 Bing introduced an AI Performance report in Bing Webmaster Tools (public preview). It shows how publisher content appears across Microsoft Copilot, AI-generated summaries in Bing and select partner integrations.

The report gives you:

• total citations
• average cited pages per day
• the “grounding queries” the AI used to retrieve your content
• page-level citation activity

Microsoft’s advice to publishers in that post:

• Use the report to see which content is already cited.
• Strengthen structure and clarity on pages that are indexed but rarely cited.
• Keep content fresh and accurate.
• Use IndexNow to notify participating search engines of changes.

It also says Bing respects content owner preferences set through robots.txt and other supported controls.

What about Claude?

Anthropic documents three bots: ClaudeBot for training, Claude-User for user-directed fetches and Claude-SearchBot for search. It says they respect “do not crawl” signals in robots.txt and that it supports the non-standard Crawl-delay extension, and it advises against blocking by IP address. Claude was not among the engines on our brief, but when you write your robots.txt you face the same three-way decision.

What happens to your clicks when AI answers appear?

Nearly every measurement shows fewer clicks on results when an AI answer is present, but the size of the drop ranges from about a third to about two thirds depending on what is measured and how. Four headline figures:

The studies measure different things, so read the middle column of the table first.

Study	What it compares	Result
Pew Research (US panel, 900 adults, March 2025, 68,879 Google searches)	Visits with an AI summary vs visits without	Clicked a result on 8% of visits with a summary vs 15% without; 1% clicked a link inside the summary; the session ended on 26% of visits with a summary vs 16% without
Ahrefs, 2025 (300,000 keywords, desktop, March 2024 vs March 2025)	Top-ranked page’s click-through rate (CTR) with vs without an AI Overview	34.5% lower CTR for position 1 when an AI Overview is present
Ahrefs, 2026 update (same sample size, December 2023 vs December 2025)	Same, over two years, against a forecast	About 58% lower CTR for position 1 than forecast (0.016 actual vs about 0.037 expected)
Seer Interactive (3,119 informational queries, 42 client organizations, June 2024 to September 2025; CTRs below are Q3 2025)	Organic CTR when an AI Overview cites you, does not cite you, or is absent	0.70% (cited), 0.52% (not cited), 1.45% (no AI Overview)
Google (Liz Reid, August 2025)	Total organic click volume from Search to websites	“Relatively stable year-over-year”; no figures, period or method published

Why do the numbers disagree?

Each one measures something different.

• Pew watches real browsing and compares visits. It covers one month in the US and is observational, and Google disputes the framing.
• Ahrefs uses search console data on 300,000 keywords, desktop only. The 58% figure replaces the earlier 34.5% because the 2026 study spans two years and uses a different baseline. Keywords with no AI Overview also lost clicks: their top-position CTR fell about 49% over the same period (0.076 to 0.039). So not all of the decline can be pinned on AI Overviews.
• Seer separates cited from not-cited. Its authors report that cited brands got roughly 35% more organic clicks, but they say causation is unproven. Its overall CTR fell in all three groups compared with a year earlier, including queries with no AI Overview.
• Google is measuring something else: total clicks across all websites, not clicks per query. Both can be true at once if there are more searches overall.

What should a small business take from this?

Being cited appears to beat not being cited, though the evidence is correlational. A citation does not guarantee a visit, either: in Pew’s data only 1% of visits included a click on a link inside the summary.

Clicks are not the only goal. The Reuters Institute’s 2026 report found that 42% of people who use chatbots for news say they always or often click through to sources, compared with 44% for search.

If your business depends on informational traffic, plan for fewer clicks per search and measure what matters to you: enquiries, sign-ups, calls, sales. If people have to come to your site to buy, an AI answer that names you can still help.

What is a practical 30-day AI SEO plan?

A realistic first month has four jobs: let the right bots in, fix what stops pages being read, rewrite key pages so they answer questions directly, and set up a way to measure results. You do not need a tool or a developer for most of it. The plan is editorial advice; where a source backs a step, it is noted.

Week 1: Check access

Week 2: Fix the basics machines trip over

Week 3: Rewrite for answers

Week 4: Earn mentions and measure

After 30 days you have crawl access decided, pages reachable, key answers written and a baseline to compare against. Our agentic navigability article explains why link structure matters beyond this checklist.

Is llms.txt worth adding?

Not as a priority. Adoption figures range widely because the studies count different things.

Study	What it counted	Adoption
Web Almanac 2025	Desktop sites	2.13%
SE Ranking, 2025	Nearly 300,000 domains	10.13%
Common Crawl, July 2026	Sampled URLs returning a text file	11.72%
Ahrefs, May 2026	Domains with traffic in Ahrefs Web Analytics, a customer base that skews technical	28%

On effect, the evidence is thin:

• SE Ranking found no correlation between having the file and being cited.
• Ahrefs found that 97% of the files it observed received zero requests in May 2026.
• Google says it will neither harm nor help Search visibility.
• Chrome’s Lighthouse agentic audit checks that the file exists, for a different purpose: readiness for browser agents.

A fetch is not proof of use, and our statistics page collects more on adoption: AI search statistics.

Which tools help with AI SEO?

AI SEO tools fall into three categories, and the right one depends on the question you need answered: “Am I mentioned?”, “Can machines read my site?” or “How do I run all my SEO in one place?” Pick by question, not by brand. We name vendors only to show what each category looks like. We have not tested them and list no prices (check each vendor’s pricing page).

The market is also moving. Adobe agreed to acquire Semrush in November 2025 (closed April 2026 per Semrush’s newsroom), and Sitecore announced in June 2026 that it was acquiring Scrunch.

Category	The question it answers	Examples (as described on vendor pages, October 2026)	Good fit
Visibility trackers	When people ask AI engines about my topic, am I mentioned and cited?	Profound, Peec AI, Otterly.AI, AthenaHQ, Goodie, Rankscale	Brands and agencies tracking many prompts over time
Site-structure and readiness analysis	Can crawlers and agents find, read and navigate my pages?	Cloudflare Agent Readiness (checks protocol items such as robots.txt, sitemap, Markdown for agents), Scrunch (AI-search visibility platform with an agent-facing content layer), page audits in some trackers	Anyone whose site is the bottleneck
Classic SEO suites with AI features	Can I keep rankings, backlinks and AI visibility in one subscription?	Semrush AI toolkit, Ahrefs Brand Radar, SE Ranking AI Search	Teams already using these suites

How to choose

• Visibility trackers work by sending a list of prompts to AI engines and recording which brands and sources appear. They tell you the score but not why. Check which engines and countries a plan covers and how many prompts you get.
• Structure analysis looks at your site itself. Cloudflare’s Agent Readiness score is protocol-based. In the pages we reviewed, we found no tool that scored a site’s internal link graph on a single number. GraphoRank does that: it scores how well AI agents can navigate, read and cite a site on a 0 to 100 scale, and it sits in this category.
• Classic suites are the easy add-on if you already pay for one.

Cloudflare’s own dataset gives a sense of where sites stand:

On demand: a Minuttia survey of 599 SEO and content professionals found 82% had implemented answer engine optimization to some degree, yet almost 40% had no dedicated budget.

If you are small, the 30-day plan costs nothing but time.

Quick answers: what should I do first?

1. Make sure the pages you care about are indexed and reachable by ordinary links.
2. Decide, bot by bot, which AI crawlers may visit, and write it in robots.txt.
3. Put the direct answer first on every key page.
4. Test your own questions by hand, monthly.
5. Treat llms.txt and special markup as optional, not as the plan.

If you want a measured starting point instead of a manual one, run a scan with GraphoRank to see how well agents can navigate and cite your site, then compare plans on the pricing page.