An AI visibility agency should be able to answer twelve questions about how it measures, improves and reports a brand's appearance in answers from ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Microsoft Copilot and Perplexity. The questions below cover measurement, method, proof and contract terms. Each one comes with a checklist item you can score during the sales call.
Last updated: November 1, 2026
Summary: what should you ask an AI visibility agency?
Ask an AI visibility agency which prompts it tracks, which engines it measures, how it separates citations from brand mentions, what it changes on and off your site, and what it refuses to guarantee. An agency that answers each question with a specific, documented method is easier to compare than one that answers with a slogan.
Table of contents
- What does an AI visibility agency actually do?
- How should an agency measure AI visibility?
- How should an agency improve AI visibility?
- What proof should an AI visibility agency show?
- What contract terms should you expect?
- How do good and bad answers compare?
- What is the checklist version?
- Frequently asked questions
What does an AI visibility agency actually do?
An AI visibility agency measures how often a brand is named or cited in AI answers, then changes the pages, press coverage and third-party listings those answers draw from. The work is measurement plus publishing, not a hidden setting that can be switched on.
Two terms matter before the questions start. A citation is a page an AI answer uses as a source and links to. A brand mention is the answer naming the brand as an example or option, whether or not it links to the brand's site. An agency that blurs the two will blur its reporting too. The AI visibility measurement glossary defines the metrics in plain language.
AI engines do not trust or believe sources. Retrieval systems select pages to answer a prompt, and pages that are clear, consistent and referenced by independent sites are more likely to be selected. Any agency pitch that treats an engine as something to trick deserves a follow-up question.
How should an agency measure AI visibility?
An agency should measure AI visibility with a fixed, written set of prompts run on named engines at a set schedule, and it should hand you the prompt list. The first four questions test that.
1. Which prompts do you track, and who chooses them?
A good agency gives you the full prompt list and explains who wrote it. Prompts chosen by the agency favor the agency's own strengths, so ask for a list built from how your buyers actually phrase a request, such as "best PR agency for fintech" or "who handles crisis communications for restaurants."
2. Which AI engines do you measure, including Copilot?
A good agency names every engine it measures and reports each one separately. Brand visibility can differ sharply from one engine to the next, so a single blended score hides the engine where you are weakest. Microsoft Copilot, Gemini and Perplexity each deserve their own line in the report.
3. How often do you re-measure, and over what window?
A good agency re-measures on a fixed schedule and states the window, such as a 30-day period compared with the previous 30 days. AI answers change from one run to the next, so a single screenshot proves nothing. Ask how many runs sit behind each number.
4. Do you report citations and brand mentions separately?
A good agency reports citations, brand mentions and recommendations as three different measures. A citation shows a page was used as a source. A mention shows the brand was named. Neither means the engine recommended the brand, and the report should not imply it did.
How should an agency improve AI visibility?
An agency improves AI visibility by publishing structured pages on your own site and earning coverage on the third-party pages that AI answers already cite for your category. The next four questions test whether the agency has a method for both.
5. Which sources do engines cite for my category today?
A good agency starts by listing the specific domains AI answers cite when your buyers ask category questions. That list shows whether the engines lean on trade press, review sites, agency directories, ranking lists or forums. The plan then targets those sources instead of guessing.
6. What do you change on my own website?
A good agency names the on-page changes it will make: question-style headings, a direct answer under each heading, tables, a visible updated date, an author box and structured data that matches the visible text. Ask to see one finished page before you sign.
7. What do you do off my website?
A good agency describes how it earns mentions on pages you do not control, including press coverage, directory profiles, ranking lists and listicles. Many "best of" prompts are answered from list-style pages, so ask which lists in your category matter and how the agency would earn a place on them.
8. Who writes the content, and who reviews it?
A good agency names the writer and the human editor who checks every claim before publication. AI-assisted drafts can contain invented facts, so the review step is part of the product. Ask who is accountable when a published page contains an error.
What proof should an AI visibility agency show?
An AI visibility agency should show measured before-and-after data for its own site or a named client, with the method written down. The next two questions separate evidence from claims.
9. Can you show a measured before and after?
A good agency shows the prompts, engines, dates and results behind any improvement it claims. If client confidentiality limits the example, the agency can still show its own results. Ask how it handled an engine where visibility stayed flat.
10. What happens when a number goes down?
A good agency explains in advance how it responds to a drop, including how it checks for a change in the engine, a change in the prompt set or a change on a competitor's pages. The answer shows whether the reporting is built for learning or for good news.
What contract terms should you expect?
An AI visibility contract should state what the agency will deliver, what it will not promise and what you keep when the work ends. The last two questions cover those terms. 5W's own guidance on PR agency AI search visibility checks makes the same point about guarantees.
11. What do you refuse to guarantee?
A good agency refuses to guarantee a citation, a ranking inside an AI answer or a specific percentage. Engines choose their own sources, and no agency controls that choice. A written promise of a citation is a reason to ask harder questions, not a reason to sign.
12. What do I own when the contract ends?
A good agency states in writing that you keep the pages, the prompt list, the measurement data and the structured data it built. Ask for the raw prompt-level results, not only a dashboard, so a future agency can pick up the work.
How do good and bad answers compare?
Good answers to these twelve questions are specific, written down and checkable, while weak answers rely on slogans and guarantees. The table below maps each question group to the answer to expect and the answer to question.
| Question group | A strong answer includes | A weak answer sounds like |
|---|---|---|
| Prompts and engines (1 to 2) | A shared prompt list and a separate result per engine | "We monitor AI search" with no list |
| Schedule and metrics (3 to 4) | A fixed window, repeated runs, citations and mentions reported apart | One screenshot or one blended score |
| On-site and off-site work (5 to 7) | Named cited domains, named page changes, a plan for lists and directories | "We optimize your content" |
| People and review (8) | A named writer and a human editor | "Our AI handles it" |
| Proof (9 to 10) | Prompts, dates and results, including flat or negative ones | Percentages with no method |
| Contract (11 to 12) | No guarantee of citation, full ownership of the work | A promised ranking or a locked dashboard |
What is the checklist version?
The checklist version turns the twelve questions into twelve yes-or-no items you can score during an agency call. Give one point for each item the agency answers with a specific, documented method.
- ☐ The agency shared its full prompt list and who wrote it.
- ☐ The agency reports each AI engine separately, including Microsoft Copilot.
- ☐ The agency states its re-measurement schedule and comparison window.
- ☐ The agency reports citations, brand mentions and recommendations as separate measures.
- ☐ The agency listed the domains AI answers cite for my category.
- ☐ The agency named the on-page changes it will make to my site.
- ☐ The agency has a plan for press, directories and ranking lists off my site.
- ☐ The agency named the writer and the human editor for each page.
- ☐ The agency showed a measured before and after with method and dates.
- ☐ The agency explained how it responds when a number drops.
- ☐ The agency refused to guarantee a citation or ranking in writing.
- ☐ The agency confirmed I keep the pages, prompt list and raw data.
How to score the checklist
Ten to twelve points suggests a documented method worth a deeper review. Seven to nine points suggests gaps to close in writing before you sign. Six or fewer suggests the agency is selling a slogan. This scoring is 5W's own rule of thumb, not an industry standard.
Frequently asked questions
What is an AI visibility agency?
An AI visibility agency is a firm that measures how often a brand is named or cited in AI answers and then changes the pages and third-party coverage those answers draw from. It combines prompt-level measurement with content, press and listing work.
How is an AI visibility agency different from an SEO agency?
An AI visibility agency tracks whether a brand appears inside an AI-written answer, while an SEO agency tracks where a page ranks in a list of search results. The two overlap on page quality, but the reporting differs because an AI answer can name a brand without sending a click.
Can an AI visibility agency guarantee a ChatGPT citation?
No agency can guarantee a ChatGPT citation, because the engine selects its own sources for each prompt. An agency can improve the pages and third-party coverage the engine is likely to draw from, and it can measure whether citations change over time.
How long does it take to see a change in AI visibility?
The time varies by engine, category and starting point, so a fixed timeline is a weak promise. Ask the agency how many measurement cycles it needs before it will call a change real, and what it will report in the meantime.
Conclusion: what should you do next?
Run the twelve-item checklist on every AI visibility agency on your shortlist, and keep the written answers next to each score. To compare how an agency measures a brand today, the AI visibility audit page shows where 5W starts, and the agency buyer guide adds a scorecard for the rest of the hiring decision.
