What is really measured
Being "visible in AI" covers several distinct things:
- Mention: does your brand appear in the answer?
- Recommendation: is it recommended, or just mentioned in passing?
- Position and tone: top or bottom of the list, positive, neutral or negative.
- Competitors mentioned instead of you.
- The sources the engine relies on to answer.
Doing the audit by hand
Write five to ten questions the way a customer would ask them, without naming your brand: a search for information, a comparison, a search for a solution to a problem, a purchase intent. Ask them to each engine in a fresh conversation, note the brands mentioned and the links provided, and repeat in a few weeks with the same questions: answers vary from one time to the next, only the trend counts.
Doing the audit with Adshift
Adshift's GEO audit generates seven questions from your keywords, industry and audience, split between transactional, comparison, problem-solving and informational intents, without naming your brand. A brand-focused mode adds questions that name it, to measure how it is perceived.
The questions go to ChatGPT and Gemini, and to Claude and Perplexity when they are available. The result gives an overall score and a score per engine, a recommendation score, the analysis by intent, sentiment, the competitors and sources cited, and the change from one audit to the next; it can be exported as a PDF. The audit costs 10 credits, refunded if fewer than 60% of the questions succeed. It can be rerun every week or every month, with an alert when the score moves beyond a chosen threshold.
What improves visibility
Engines recommend what they can identify and justify. Three workstreams matter, without depending on other websites:
- A clear identity: the same definition sentence (what you are, for whom, where) on your website, in your structured data and in an llms.txt file.
- Pages that directly answer the questions where you are missing: short answer first, verifiable facts, update date.
- Honest comparisons with the competitors AI engines mention, in your real category, saying when the other option is a better fit.
The action plan
After the audit, Adshift generates an action plan for 5 credits: six to eight recommendations ranked by priority, each tied to a question, engine or source from the audit. The plan does not promise figures: when a number would be useful, it asks you to measure it rather than make it up.
Frequently asked questions
- Do AI answers change from one time to the next?
- Yes. The same question can get different answers depending on the moment and the engine. That is why an audit asks several questions to several engines and is repeated over time: you follow a trend, not an isolated answer.
- How long does it take to appear in AI answers?
- It depends on the engine. Those that search the web when answering, such as Perplexity, can pick up a new page within a few weeks once it is indexed. A model's own knowledge changes more slowly, at the pace of its updates.
- Should you name your brand in the audit questions?
- Not to measure visibility: a customer who does not know you will not name you. Yes to measure brand perception, which is what Adshift's brand-focused mode does.