How to get your business recommended by ChatGPT

By The SignalUp Team · Updated August 4, 2026

When a buyer asks ChatGPT "who should I use for X," the answer names a handful of brands. Those recommendations aren't random — they're a function of what the model learned in training, what its search tools retrieve at answer time, and how confidently it can describe each brand. All three are influenceable. The same steps apply to Claude, Gemini, Perplexity, and Copilot.

1. Make your site readable without JavaScript

This is the step most businesses fail without knowing it. GPTBot, ClaudeBot, PerplexityBot and Amazonbot do not execute JavaScript. If your site is a client-rendered app, these crawlers see an empty shell. Serve real HTML — server-rendered, statically generated, or prerendered at build time — for every public page. Test it: curl your homepage and check that your actual copy appears in the response.

2. Welcome AI crawlers in robots.txt

Explicitly allow GPTBot, OAI-SearchBot (ChatGPT's live-search crawler), ClaudeBot, PerplexityBot, Google-Extended, Amazonbot and Bingbot (Copilot runs on Bing's index) on marketing pages. Keep private areas disallowed.

3. Publish llms.txt

An llms.txt file is a concise, machine-readable summary of what you do, who it's for, what it costs, and where your key pages live. It gives every engine a canonical description of your business to draw from instead of guessing.

4. Add JSON-LD structured data

Organization, Service or Product with Offer prices, and FAQPage schema remove ambiguity. An engine that can read "Hold, $590/month, cancel anytime" as structured data will describe you more accurately — and more confidently — than one parsing prose.

5. Make your claims quotable

Engines lift concrete, single-sentence facts. "Free AI visibility scan, no credit card, results in minutes" gets quoted; three paragraphs of brand storytelling does not. Audit your key pages: can every important claim be extracted as one standalone sentence?

6. Earn corroborating mentions

Ask Perplexity a buying question in your category and note which sources it cites — review sites, directories, comparison posts, industry publications. Those are the pages the engines already trust. Being present there (accurately described, consistently named) is the off-site half of AI visibility, and it also feeds the training data of the next model generation.

7. Measure, then repair on a schedule

You can't manage what you don't measure. Ask the engines the questions your buyers actually ask, on a schedule, and track your Share of Voice against competitors. Answers drift as models retrain and indexes refresh — visibility you won in March can quietly vanish by June. Re-measure weekly and repair when it slips.

How long does it take?

Retrieval-augmented engines (Perplexity, ChatGPT with search, Copilot, AI Overviews) can reflect on-site fixes within days to weeks of recrawling. Pure training-data visibility moves on model-release timescales — months. That's why the practical playbook is: fix the retrieval layer now, build the corroboration layer continuously.