How to check whether your business appears in ChatGPT answers, and what to do if it does not

Spend twenty minutes measuring first

Write ten to fifteen questions your customer asks immediately before buying. Not questions about your company name, but about your category: “best Italian restaurant”, “reliable shipping company near me”, “how much does X cost”. These are the questions where deals are decided.

Ask each question in ChatGPT, Gemini and Perplexity, then repeat it in Arabic. Record it in a simple table: question, engine, language, who appeared. Two things become obvious immediately: these engines typically name three businesses rather than ten, and Arabic results differ substantially from English ones.

Repeat each question at least twice in separate sessions. These engines are non-deterministic, so the same question can produce two different answers. Anyone showing you a single fixed “ChatGPT ranking” number does not understand how they work.

Cause one: no answer can be lifted from your content

These engines build answers by lifting passages from sources and attributing them. If your page is written like a brochure (“we are pioneers in delivering innovative solutions”), there is no sentence in it that can be lifted and presented as an answer.

The fix is simple in principle: open every page and every section with a direct answer in the first two lines, using specific names and numbers. “We deliver the audit in seven business days at a fixed price of X” is extractable. “We are committed to the highest standards of quality” is not.

This pays off twice: the same structure that makes content quotable also makes it clearer for a human reader. I have not seen a case where the two requirements conflicted.

Cause two: your name is missing from the sources

These engines do not invent names or recommend from nothing. They quote specific sites: business directories, local press, category platforms, and the “best of” round-ups. If your name is not on those sources, it will not appear in the answer no matter how well you optimise your own site.

This part is the slowest and hardest, and also the most durable. Building real mentions on real sources usually takes two to three months to show up, but unlike the other two, it does not collapse when an algorithm updates.

What does not work: fake reviews and planted forum threads. They get detected, and when they do they take your genuine credibility with them.

Cause three: you are blocking them without knowing

This is the most frustrating cause and the easiest to fix. Open robots.txt on your domain and look for GPTBot, PerplexityBot, ClaudeBot and Google-Extended. If you find a Disallow against any of them, you are preventing that engine from reading you.

Many of these rules were never added deliberately. They arrive as a CDN default or inside a ready-made template. Check your firewall too: permission in robots.txt means nothing if a WAF returns 403 before the request lands.

There is a real commercial judgement here: some publishers block these crawlers deliberately to protect their content. If you sell a service or a product you almost certainly want the opposite: you want to be quoted.

And why llms.txt is not the answer

You will find people selling llms.txt as the key to AI visibility. There is no evidence for that. Large-scale analysis has found no correlation between having the file and being cited, and Google itself has indicated it is unnecessary.

Add it if you like, since it costs nothing and does no harm. But do not build an expectation on it or pay much for it. The same applies to structured data: good technical hygiene, not a cause of inclusion in Google’s AI Overviews.

The working rule: AI visibility is roughly 80% strategy, credibility and sources, and 20% technical. Anyone selling you the reverse is selling you the easy part, because it is the part they can deliver in an hour.

If you would rather have this measurement done for you than write the questions, run all three engines, and weigh the results by hand, that is exactly one of the things the Digital Presence Audit checks: whether you actually appear in AI answers, not whether you might.