How visible is your brand in ChatGPT recommendations?

A branded prompt can test whether ChatGPT understands your company. An unbranded buying prompt tests whether you are part of the market shortlist. Those are different jobs. This calculator focuses on the second job because that is where new demand can be won or lost.

OpenAI publishes crawler documentation, but no public calculator can know every signal behind a ChatGPT answer. The practical measurement surface is the answer itself: whether you are named, whether competitors are named, what evidence appears, and whether facts are accurate.

That is why a useful score keeps the commercial context visible. A recommendation for the wrong buyer, wrong location, or wrong feature set should not be counted like a clean recommendation or presented as a win.

Good prompts include a category, audience, and buying context. For example: best CRM for boutique law firms, top Webflow agencies for B2B SaaS, or which payroll tool should a 50-person startup use. The sharper the buyer situation, the more useful the output.

Do not overfit one perfect prompt. A real buyer asks in several ways. Use a small set of variants around the same intent, then count whether your brand appears consistently enough to trust the pattern.

Direct recommendations carry the most weight because they place the brand in the strongest buyer position. Shortlist mentions count too, but they are weaker. Owned source exposure helps because it suggests ChatGPT can reach or reference evidence controlled by the brand.

Competitor recommendations and wrong facts reduce the score. A competitor-heavy answer means the buyer's consideration set is moving away from you. A wrong fact means the answer may be visible but commercially unsafe.

Do not obsess over the exact number. The band matters more because it tells you what kind of work comes next. Severe gaps usually need source discovery and competitor analysis. Uneven scores usually need targeted fixes. Strong scores need monitoring and expansion into adjacent prompts.

For local service firms, geography can change the band quickly. For B2B SaaS, use case and buyer role often matter more than location. The score should be interpreted inside the market context, not as a universal benchmark.

One ChatGPT answer can be interesting, but it is not a measurement system. Run prompts in a consistent environment, preserve the text, and retest when fixes go live. That gives you a trend instead of a screenshot.

Another distortion is treating citations as complete transparency. Visible citations can be useful, but they may not reveal every influence behind the answer. That is why the calculator includes both source exposure and repeat stability rather than relying on citations alone.

The fastest wins often come from boring fixes. Make the product, service, pricing context, integrations, limitations, locations, and use cases explicit on crawlable pages. Add FAQ structure where it genuinely helps users. Fix stale directory and marketplace profiles.

Then address competitor framing. If ChatGPT repeatedly recommends a competitor because their comparison page explains the category better, create a fair, specific, evidence-rich comparison page. The point is not to trick the model. The point is to make the public evidence more complete.