AI SEO is the work of making your site visible in AI search: ChatGPT, Perplexity, Claude, Microsoft Copilot and Google's AI Overviews and AI Mode. In practice it means three kinds of change on your site: open access for the AI crawlers that cite you, serve your text in a form they can read, and state your facts clearly enough to be quoted. This guide is the hands-on half of our guide to generative engine optimization, which maps the whole topic: here you get the changes themselves, in order, with what each operator actually documents. Expect the technical part to take an afternoon; the content part is ongoing.
Quick answer
- AI SEO is SEO with extra switches, not a new discipline. Google says optimising for its generative AI features is still SEO, and most AI engines retrieve pages through search indexes.
- Allow the AI search crawlers: OAI-SearchBot, Claude-SearchBot, PerplexityBot. OpenAI documents that blocking OAI-SearchBot keeps you out of ChatGPT search answers. Training bots such as GPTBot are a separate switch.
- Serve your main text in the HTML. The major AI crawlers do not run JavaScript, so content that appears only after scripts is invisible to them.
- Be indexed in Google and Bing. AI Overviews draw on Google's index; Copilot grounds answers in Bing; ChatGPT search uses third-party search providers.
- Write answer-first pages: the fact in the first sentence, numbers with units, an author and a date.
- Schema helps some engines and llms.txt is a 15-minute bet; neither is required by any operator.
- Check your side in a minute with the free AI crawler check.
One discipline, several names
You will see this work sold as AI SEO, GEO (generative engine optimization), AEO (answer engine optimization), LLM SEO and AI search optimization. They are the same discipline under different labels: none has an official definition, and Google's guide to its generative AI features says that, from Google Search's perspective, all of it is still SEO. The labels differ mostly in where they came from (AEO grew out of featured snippets and voice search, GEO out of a 2023 research paper) and our comparison of answer engine optimization vs GEO vs SEO traces that history. For this guide the name does not matter; the changes below do.
What every label points at is the same chain: a page must be allowed (the crawler may fetch it), readable (the text is there without JavaScript), retrievable (it ranks in an index the engine searches) and quotable (it states an answer clearly enough to be used). The steps follow that chain.
Step 1: let the AI search crawlers in
Every major operator runs separate crawlers for training and for AI search, and the search ones are the crawlers that cite and link you. OpenAI documents that sites which disallow OAI-SearchBot are not shown in ChatGPT search answers; Anthropic documents Claude-SearchBot and Perplexity documents PerplexityBot the same way. Blocking a training bot such as GPTBot or ClaudeBot does not affect citations, so you can stay out of training and still be cited, but only if you keep the groups separate:
# robots.txt: out of model training, visible in AI search
User-agent: GPTBot
User-agent: ClaudeBot
User-agent: CCBot
Disallow: /
User-agent: *
Allow: /
Sitemap: https://example.com/sitemap.xmlCheck two layers beyond robots.txt. Cloudflare's AI bot blocking, security plugins and hosting firewalls can refuse a crawler that robots.txt allows, and a stray Disallow: / under User-agent: * takes Googlebot down with the AI bots. AI crawlers and robots.txt has complete files for four kinds of site, the group rules that trip people up, and how to enforce a real block at the edge.
Step 2: serve your text without JavaScript
An analysis of AI crawler traffic by Vercel and MERJ, published in December 2024, found that none of the major AI crawlers rendered JavaScript: OpenAI's, Anthropic's and Perplexity's included. They read the HTML your server sends, and some fetch script files only as text. If your prices, product descriptions or answers appear only after scripts run, those crawlers see none of it.
The fix is server-side rendering or static generation for everything you want quoted. Menus and widgets can stay client-side; body text, prices and answers cannot. Watch the smaller leaks too: prices fetched from an API after load, reviews injected by a widget, JSON-LD added by a tag manager. Test it by viewing the page source (not the inspector) and searching for a sentence you want cited.
Step 3: be indexed where AI engines look
AI answers are built from pages retrieved through search indexes, so classic indexing is the cheapest AI search optimization there is:
- Google: AI Overviews and AI Mode use Google's index, and Google says a page needs nothing beyond being indexed and eligible for a snippet.
- Bing: Copilot grounds its answers in Bing, and Bing has been ChatGPT's announced search partner since 2023. Verify the site in Bing Webmaster Tools (most sites never have) and submit the sitemap.
- IndexNow: if your CMS or CDN supports it, Bing hears about new and changed pages at once.
- The basics: no stray
noindex, correct canonicals, a sitemap listed in robots.txt, no 5xx answers under load.
Step 4: structure content so answers can quote it
AI answers are assembled from sentences that state something plainly. No operator documents a writing style as a ranking rule (Google explicitly says you do not need to write differently for AI) but what gets quoted is consistent, and it is the same thing that earns featured snippets:
- Answer first. Put the one-sentence answer directly under the heading, then explain. "Delivery to Germany takes 3-5 working days and costs 9 EUR" is quotable; "we pride ourselves on fast shipping" is not.
- Numbers with units. Prices with currency, sizes in centimetres, times in days, dates on everything.
- One question per section, phrased the way a buyer asks it, with the follow-ups (cost, time, alternatives) on the same page.
- Say who wrote it and when. An author, an updated date and a way to contact you help any reader, human or model, trust the page.
- Keep your facts consistent everywhere. Name, address, prices and opening hours should read the same on your site, your Google Business Profile and directories; conflicting facts give an answer a reason to cite someone else.
- Cover what only you know: your prices, your process, your data. Pages that repeat what every competitor says give an answer nothing to cite you for.
Step 5: state your facts in schema
Structured data states price, availability, opening hours and who runs the business in a form machines read directly. The honest record: Google says it is not required for its AI features, Microsoft says schema helps its language models understand content, and OpenAI says ChatGPT's shopping results use structured metadata such as price and description. Nobody documents it as a citation factor, and there is no special "AI schema" type.
Add it anyway: it is cheap, it powers rich results in classic search, and it costs nothing to AI. Put JSON-LD in the server's HTML rather than a tag manager, keep it matching the visible page, and start with Organization (with logo and sameAs), Product with Offer, LocalBusiness and Article. Schema markup for AI search lays out each operator's statement with its source and the types worth adding.
Step 6: add an llms.txt, as a cheap bet
llms.txt is a Markdown file at /llms.txt that gives assistants a curated map of your site. Be realistic: it is a proposal, not a standard; Google says Search ignores it, and no major AI search engine has committed to reading it. Coding agents and documentation tools do use it, and it takes about 15 minutes, so it is a reasonable extra; never the strategy. llms.txt: what it is and how to write one covers the format and placement, and the generator drafts one from your sitemap:
Per-engine notes
The steps above serve every engine; what differs per engine is small and documented:
| Engine | What gets you in | First thing to check |
|---|---|---|
| ChatGPT search | OAI-SearchBot allowed; found via third-party search providers | robots.txt and firewall allow OAI-SearchBot |
| Google AI Overviews / AI Mode | Indexed and snippet-eligible in Google Search | Page indexed; no nosnippet you did not mean |
| Microsoft Copilot | In Bing's index | Verified in Bing Webmaster Tools, sitemap submitted |
| Perplexity | PerplexityBot allowed | CDN or firewall not blocking it |
| Claude | Claude-SearchBot allowed | Separate from ClaudeBot, the training crawler |
For the two biggest, we have step-by-step guides: SEO for ChatGPT covers OpenAI's three crawlers, Bing indexing and product feeds, and how to rank in Google AI Overviews covers eligibility, snippet controls and how pages are picked.
Check your site's side of it
The check reads your robots.txt for 18 AI crawlers grouped by purpose (training, AI search, user fetches) validates llms.txt, looks for AI opt-out tags and compares the raw HTML with the rendered page, so you see steps 1, 2 and 6 answered for your site in one run. Every getReport report includes the same findings.
Measuring what it changes
Improvement shows up as AI search visibility: how often AI answers name your brand, cite your pages or send you visitors. Since 2026 some of it is properly measurable (Search Console has a Generative AI performance report, Bing Webmaster Tools counts Copilot citations per page, and your analytics and server logs show the rest) and much of it still is not. What AI visibility is and how to measure it covers every source, its limits and a baseline you can repeat monthly; this page stays about the changes you make.
Common mistakes
- Blocking every AI bot, then wondering why no AI cites you. Training, search and user fetches are separate switches; decide per purpose.
- Leaving your facts in JavaScript. A page whose prices render client-side is empty to most AI crawlers, whatever else you optimise.
- Ignoring Bing. It feeds Copilot and has fed ChatGPT search since 2023; verifying it takes minutes.
- Buying "AI schema" or treating llms.txt as the strategy. No such schema type exists, and no AI search engine has committed to llms.txt. Both are extras on top of access, rendering and content.
- Rewriting the whole site "for AI". Google says there is no need. Fix access and rendering, then improve the pages people actually ask about.
Questions people ask
What is LLM SEO?
LLM SEO is another name for AI SEO: optimising a site so products built on large language models (ChatGPT, Perplexity, Claude, Copilot, Google's AI features) read and cite it. The work is the same under every label: allow the AI search crawlers, serve your text in plain HTML, be indexed in Google and Bing, and publish pages that answer questions directly. No operator uses the term officially.
Is AI SEO different from traditional SEO?
Only at the edges. The foundation (crawlable, indexable, genuinely useful pages) is identical, and Google describes optimising for its AI features as still SEO. What is new is small: each AI operator runs its own crawlers with separate robots.txt switches, most of those crawlers do not run JavaScript, and success means being cited in an answer rather than ranking in a list.
Does AI SEO replace Google SEO?
No. AI engines retrieve pages through search indexes, so ranking in Google and Bing is what gets you retrieved: AI Overviews are built from Google's index, Copilot grounds answers in Bing, and ChatGPT search uses third-party search providers alongside OpenAI's own crawler. Dropping classic SEO would remove you from AI answers too. They are mostly the same work; do it once for both.
Do I need an AI SEO agency or a special tool?
Usually not. The documented work (robots.txt, server-rendered text, indexing in Google and Bing, clearer pages) takes hours and can be checked for free. Before paying for more, ask what a service measures, which engines it covers and how much its numbers change between runs, because no vendor has access to internal AI metrics. Trackers can show trends for a large brand; they are not a requirement.
Which AI search engines should I optimize for first?
Start with the largest audiences and the clearest rules: Google's AI Overviews and AI Mode, which follow normal Google SEO; ChatGPT search, which needs OAI-SearchBot allowed and good Bing indexing; and Microsoft Copilot, which answers from Bing. Perplexity and Claude follow the same pattern with their own crawlers. The fixes overlap almost completely, so opening access once serves all of them.
How long does it take to improve visibility in AI answers?
It depends on the source. OpenAI says robots.txt changes for its crawlers apply within about 24 hours, and search-based answers pick up new or updated pages once they are crawled and indexed, often within days or weeks. What a model learned in training changes only when a new model is trained, which takes months, and you cannot speed that up.