1. Baseline what the platforms say today
Write the questions your buyers would actually type — not keywords, questions — and run them across ChatGPT, Gemini, Perplexity, Copilot, and AI Overviews. Record whether you appear, what is said, and which sources are cited instead.
Without a baseline there is nothing to improve against, because model outputs vary between runs and a single lucky answer proves nothing.
2. Make your entity unambiguous
Models resolve a brand to an entity before they say anything about it. Conflicting names, categories, and descriptions across your site, directories, and profiles make that resolution fail, and a model that is unsure will simply name a competitor it is sure about.
- One legal name, one trading name, used consistently.
- The same one-sentence description everywhere it appears.
- Consistent category, location, and contact details across every profile.
3. Let the retrieval bots in
Check robots.txt, WAF rules, and bot protection for OAI-SearchBot, ChatGPT-User, PerplexityBot, and Google-Extended. Then confirm the content you care about is in the server-rendered HTML rather than injected by client-side JavaScript after load.
4. Write answer-shaped content
Models quote passages, not pages. Lead with the direct answer in the first sentence, then the specifics underneath, and keep each section self-contained so it still makes sense lifted out of context.
- One question per heading, phrased the way a buyer asks it.
- Concrete numbers, names, and dates rather than adjectives.
- Short paragraphs; no answer buried under 400 words of preamble.
5. Add structured data and an llms.txt
Schema markup gives machines the same facts your page gives humans, without ambiguity: Organization for the entity, Service for what you sell, FAQPage for questions, Article for writing, BreadcrumbList for context. An llms.txt file summarises the site in plain text for language models.
6. Earn the citations models already trust
Your own site is evidence of what you claim; third-party sources are evidence that it is true. Look at which domains the platforms currently cite for your category — review platforms, directories, trade press, comparison pages — and get accurate, complete entries on those specific sources rather than chasing generic links.
7. Re-run the prompt set and iterate
Re-run the same prompts monthly against the same schedule. Track mention rate, position, sentiment, and cited sources per platform, then spend the next month on the prompts where you are still absent. Improvement in AI visibility is a compounding loop, not a one-off project.