Blog

How to tell whether ChatGPT recommends your brand and which page to improve first

Run an AI visibility audit with real buyer questions. Track mentions, citations and visits separately, then choose a change you can measure again.

Take a real sales question that leaves out your company name. Ask it in ChatGPT with web search enabled and save the answer and its sources. Record whether the brand appears, whether the answer recommends it for the job and whether it links to a page you own. Each observation points to a different next step.

Which questions belong in an AI visibility audit?

Ask sales for five questions buyers recently asked before hiring. Keep the details that change the answer: location, type of buyer, existing technology and constraints. A branded question checks what the system says about you. For a discovery test, describe the problem the way a buyer would before learning your name.

For a web studio in Mexico, try three stages of a purchase:

  • Discover providers. “Which agencies in Mexico can migrate a bilingual WordPress site to Webflow?”
  • Compare options. “We have two editors and 150 technical product pages. What should we ask agencies for to compare proposals?”
  • Verify a provider. “What public evidence shows that MANUBRIO builds Webflow sites its clients can edit?”

The comparison question can reveal an editorial opportunity even if the answer names no agencies. Open the guides it cites and identify the buyer question each one actually answers.

How to record a result and interpret it

Start a new conversation for each test. Save the exact question, date, tool, language, requested market and search mode. Copy the passage with the recommendation and open the cited URL. A partner directory that describes your studio can support the recommendation while receiving the potential visit itself.

A fictional row might read: “Bilingual migration · brand mentioned: yes · recommended for this job: yes · own domain cited: no · source: partner profile · missing evidence: bilingual case study.” Use a separate row when repeating the exercise in Gemini or Perplexity. Record the answers in which your company is absent as well.

Count distinct questions in each round. If three of ten mention you, report “three of these ten questions.” The sample measures what happened in that exercise. Save it so the team can run the same test after making a change.

Which page should you improve?

The source misstates your service. Open the cited page first. If you own it, explain who the service is for, what you deliver and where the scope ends. If the source is a partner profile your team manages, reconcile its description with the website.

Buyers ask for experience and find a feature list. Publish a case describing the starting problem, one implementation decision, the deliverable and evidence you have permission to share. For a bilingual migration, show an example record and the task its editor can now complete independently.

The right page exists but is hard to find. Link it from the relevant service, check indexing and make the main explanation available as readable text. Google applies its core SEO guidance to its AI features. OpenAI describes the role of OAI-SearchBot in discovering content for ChatGPT search. Google Search Central · OpenAI publisher FAQs

How to check what changed

Log the page, the buyer question and the publication date. Repeat the round under similar conditions and open the sources again. Track referred visits and enquiries separately. A citation can appear without a click, and a prospect may call days later. Asking how they found you helps recover some of that context.

To start a visibility audit with MANUBRIO, bring three buyer questions and an answer that recommends someone else for work you also do. That gives us a concrete page to investigate. Explore an AI visibility audit

← Back to the blog

Start a project

Let's build
what's next.

Tell us about your business and what you want to achieve. We'll recommend the right stack and a clear plan, no strings attached.