AI assistants don't invent their answers — they search, read, and cite. Which means you can influence it. Five practical levers, most of them ten-minute jobs.
ChatGPT's search and Microsoft Copilot retrieve live results from Bing's index. Perplexity and Claude combine classic search with their own crawlers. The assistant then summarises the best sources — and cites them. So AI visibility stands on two legs: be findable in the indexes they use, and be the page whose answer is easiest to quote. Regular SEO carries most of it; the five levers below cover the rest.
ChatGPT, Copilot and Perplexity read the web with their own crawlers (GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot). If your robots.txt blocks them — some templates and "privacy" presets do — you simply don't exist in AI answers. Blocking them saves you nothing and costs you every AI recommendation.
Check my site's robots.txt. Make sure it does NOT block these user-agents: GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, bingbot. If any of them are disallowed, remove those blocks. Keep any other rules as they are, and keep the Sitemap line. Show me the final robots.txt.
Here's the thing almost nobody does: ChatGPT's search and Microsoft Copilot answer from Bing's index. Everyone registers with Google Search Console; almost nobody registers with Bing — so this is a free head start. Go to bing.com/webmasters, sign in, and use "Import from Google Search Console" — it copies your verification and sitemap in one click. Ten minutes, done forever.
AI assistants lift direct answers: a clear definition early on the page, FAQ sections with real questions, lists and steps, and FAQ/Article schema markup. A page that answers the question in its first paragraph gets cited; a page that warms up for four paragraphs gets skipped.
Review my most important pages and make each one "quotable" for AI assistants: answer the page's core question directly in the first paragraph, add an FAQ section with 3-4 real questions and concise answers, and add FAQPage schema markup (JSON-LD) matching those questions. Keep the tone natural — no keyword stuffing.
llms.txt is an emerging convention — robots.txt's friendly cousin. A small text file at yoursite.com/llms.txt that tells AI models what your site is, who it's for, and which pages matter. It costs nothing, takes five minutes, and early adopters get the benefit while it's still rare.
Create an llms.txt file at my site's root (served at /llms.txt as plain text). Format: a # heading with the site name, a one-paragraph plain-language description of what the site offers and who it's for, then a "## Key pages" list with the most important URLs and a one-line description of each. Write it in the site's own language, keep it under 40 lines, and make sure it's served as text (not HTML).
Perplexity and ChatGPT weigh discussions and reviews heavily — Reddit, forums, comparison articles. You can't fake this honestly (and astroturfing gets brands banned), but you can earn it: answer questions in your niche's communities as yourself, get listed in genuine comparison posts, and make your product worth mentioning. Slow, real, and the strongest signal of all.
The free scan now checks whether AI crawlers can read your site and whether you have an llms.txt — alongside the classic SEO foundations.
ChatGPT's search (and Microsoft Copilot) retrieve live results from Bing's index, then the model summarises and cites what best answers the question. So the recipe is: be indexed by Bing, rank for the question, and structure your page so the answer is easy to lift and cite.
About 80% is the same: crawlable site, indexed pages, content that genuinely answers searches. The extra 20%: allow AI crawlers in robots.txt, register with Bing (not just Google), structure content in quotable answers with FAQ schema, and consider an llms.txt file.
For a business site that wants customers: almost certainly not. Blocking AI crawlers doesn't stop competitors — it just removes you from the answers your future customers read. Save blocking for content you truly don't want quoted.
It's early — no assistant guarantees it's read, but several AI crawlers fetch it, it costs five minutes, and it can only help. Think of it like submitting a sitemap in 2006: cheap insurance on where things are heading.
Ask them: search ChatGPT, Copilot and Perplexity for the questions your customers ask and see if you're cited. Watch your analytics for referrals from chatgpt.com and perplexity.ai. (Automatic AI-visibility tracking is on OptiSearch's roadmap.)