Be the source AI answers cite
AI Overviews, AI Mode, ChatGPT search, Perplexity and Copilot answer first and cite second. AI search engine optimization is the work of making your site the one they retrieve, trust and quote, and measuring it every week. It is built on classic SEO, because the assistants still start from search indexes, and it changes the scoreboard from positions to citations.
Get retrieved
Crawler access for OAI-SearchBot, ChatGPT-User, PerplexityBot and ClaudeBot; Bing indexing (ChatGPT search overlaps Bing's top results far more than Google's); IndexNow; fast, clean HTML. We publish llms.txt too, while being clear that Google says it does not use it.
Get trusted
Consistent entity data everywhere (Organization and LocalBusiness schema, sameAs, name, address and phone), named authors, dated pages, cited sources, brand mentions from digital PR.
Get quoted
Answer-first paragraphs, question-form headings, definitions and tables assistants can lift verbatim, original data nobody else has.
What changed, and why it matters to your traffic
Two things happened at once. Google started answering many informational queries on the results page with an AI Overview, and a meaningful share of research moved to assistants. Ahrefs measured a 58% lower average click-through rate for top-ranking pages on queries with an AI Overview, and Semrush found a similar drop across 17.8 billion queries. Informational pages that lived on clicks lose them; pages that are cited inside the answer keep a share of them; and commercial queries, where people still need to choose a supplier, are less affected. That is why the work below is about being cited rather than only ranked, and why we measure both.
Where each assistant gets its answers
| Surface | Retrieves from | What that means for you |
|---|---|---|
| Google AI Overviews and AI Mode | Google's own index | Classic Google SEO is the prerequisite; Search Console reports AI impressions per page |
| ChatGPT search | Largely Bing's index, plus its own OAI-SearchBot crawls and live ChatGPT-User fetches | Bing Webmaster Tools, IndexNow and unblocked OpenAI crawlers matter more than most sites realise |
| Perplexity | Its own crawler, PerplexityBot, and live fetches | Blocking the bot in robots.txt or at the CDN removes you entirely |
| Microsoft Copilot | Bing's index | Same as ChatGPT search: Bing indexing and freshness |
| Claude | Its own crawler, ClaudeBot, and live fetches | Same access rule; check CDN bot-protection defaults |
What we do, month by month
- Baseline: a 40-prompt panel per topic across Google AI Overviews, ChatGPT, Perplexity and Copilot; who is cited today and why.
- Access: robots.txt and firewall rules (CDNs often block AI agents by default), Bing Webmaster Tools, IndexNow, sitemap hygiene; llms.txt as a courtesy, not a ranking signal.
- Entity work: schema, sameAs, profiles, consistent descriptions of who you are across the web, so every source the assistants read agrees.
- Answer-first rewrites: the pages that should be cited get openings that answer, headings that ask, and facts that can be quoted with a date and a source.
- Mentions: data-led assets pitched to publications and communities the assistants already trust. Ahrefs' study of 75,000 brands found unlinked brand mentions correlate with AI citations about three times more strongly than backlinks.
- Measurement: citation share, AI referrals and leads in your monthly report, with the panel re-run on a schedule.
What it can and cannot achieve
It can get a site that was invisible to assistants retrieved within weeks, because the access problems are mechanical. It can make your pages the ones an assistant quotes when it does cite a source in your topic, if the content is the clearest and best-supported answer available. It can turn AI visits into measurable leads.
It cannot guarantee a citation for a given prompt, because answers are assembled per query and per session and change between runs. It cannot recover the clicks an AI Overview removes from a purely informational query; the honest strategy there is to be the cited source and to build pages for the commercial questions that still bring visits. And it cannot substitute for authority: assistants favour sources that are already well described and mentioned across the web.
What we can and cannot measure
- Measured: AI referral sessions and the leads they produce, in GA4; which pages assistants fetched and how often, in server logs; citation share on a fixed prompt panel over time; AI Overview and AI Mode impressions per page, in Search Console.
- Inferred: clicks from AI Overviews (Google does not report them) and the effect of a citation on brand searches.
- Not measurable by anyone: a guaranteed placement in a specific future answer. Any tool or agency claiming it is describing a sample as a certainty.
What we will not claim
Nobody can guarantee a citation, and the assistants change weekly. We do not sell llms.txt as a ranking signal, we do not report a single prompt run as proof of ongoing visibility, and we do not forecast AI traffic to a number. What we show is the panel moving, the referrals arriving and the leads that follow, with dates. Our AI search optimization service page describes the deliverables and which plan includes what, and the 2026 guide goes deeper on each step.
Questions
What is AI search optimization?
Making a website visible and quotable in AI-generated answers: Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Microsoft Copilot and Claude. It combines classic SEO (the AI systems still retrieve from search indexes) with answer-first content, entity consistency, crawler access, structured data and citation tracking.
Is it different from SEO?
It is SEO with a different scoreboard. Rankings still matter because retrieval starts from indexes (Google and Bing). What changes is the emphasis: being the clearest, most quotable, most consistently described source, and measuring citations and AI referrals rather than positions alone.
Should we block AI crawlers?
Blocking training bots (GPTBot, Google-Extended) is a policy choice. Blocking the search and retrieval agents (OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot) removes you from the answers your customers read. We recommend allowing retrieval and deciding separately on training.
Does llms.txt help?
Google has said it does not use llms.txt for Search or AI Overviews, and no major assistant has committed to it publicly. We publish one because it costs nothing and some tools read it, but we do not count it as a ranking or citation signal and we will not sell it as one.
How do you measure it?
A fixed prompt panel per topic re-run on a schedule across the assistants, citation share over time, AI referral sessions in GA4 (chatgpt.com, perplexity.ai, copilot.microsoft.com referrers and utm_source=chatgpt.com), crawler activity in server logs, and leads attributed to those sessions. Google reports AI Overview impressions in Search Console but not clicks, so that surface is tracked by impressions and inferred.
How long does it take?
Access and entity fixes can change what assistants retrieve within weeks, because the retrieval crawlers visit often. Being cited consistently for competitive topics takes months of content and brand-mention work, and the answers keep changing, so we report a trend rather than a date.
Read more
Find out who AI cites for your topics.
We run the prompt panel before the call and show you the gap.