AI answer engines cite pages they can reach, read and trust. This checklist covers the technical access first, then content and trust. Download it as a spreadsheet with a status column, or tick it off here.
Generative engine optimization (GEO) is young, and nobody outside the AI companies knows exactly how sources are chosen. Everything below is either documented by the vendors or good practice that also helps classic search.
Access
- OAI-SearchBot is allowed in robots.txt — needed for ChatGPT search; check with the AI crawler checker
- PerplexityBot and Claude-SearchBot are allowed — search crawlers for Perplexity and Claude answers
- Googlebot and Bingbot are allowed — AI Overviews use Google's index; Copilot uses Bing
- Training crawlers (GPTBot, ClaudeBot, CCBot, Google-Extended) are allowed or blocked on purpose — a separate decision from search
- The firewall and CDN do not block verified AI search crawlers — bot protection can override robots.txt
- Key pages answer 200 without login, cookie walls or geo-blocks — test from outside your network
Readable content
- The main text is in the HTML source, not only after JavaScript runs — view source, search for a sentence from the page
- Each page has one clear H1 and descriptive H2 headings — headings that match how people ask
- Prices, specifications and key facts are text, not images — models cannot read text inside images reliably
- Tables are real HTML tables — for comparisons and specifications
- No key content hidden in tabs that load on click — content loaded only on interaction may be missed
Clear answers
- Each key page answers its main question in the first two sentences — then adds the detail
- Pages cover one topic each, without near-duplicates — duplicates split the answer
- Facts are specific: numbers, dates, conditions, names — vague copy gives nothing to quote
- An FAQ answers the questions customers actually ask — visible on the page, not only in markup
Trust signals
- An About page says who runs the site — with real names or the company
- Contact details are easy to find — address, e-mail or phone
- Articles show an author and an update date — and the date is real
- Organization structured data with name, logo and official profiles (sameAs) — helps tie the site to your brand
- Business details are consistent across the web — name, address, phone the same everywhere
Optional extras
- An llms.txt lists the key pages — low cost, limited adoption; check it with the llms.txt checker
- The site is verified in Bing Webmaster Tools — ChatGPT search also uses third-party search providers
Measure
- Ask ChatGPT, Perplexity and Google the questions your pages answer and note who is cited — repeat monthly
- Track referrals from chatgpt.com, perplexity.ai and other AI tools in analytics — as their own channel
- Review Search Console for pages that lost clicks to AI Overviews — compare impressions and clicks over time