How to choose an AI Visibility provider: seven questions worth asking

The AI Visibility services market currently looks much like the SEO market did in 2010. Plenty of firms claiming they can "optimise for AI". Few standards. Even less transparency about methodology. And, hardest of all, results that are difficult to compare, because everyone uses different metrics or shows none at all.

The point is not to distrust every provider - it is to ask the right questions before signing. Below are seven questions that separate firms with a real methodology from those selling general consulting under a fashionable label.

Question 1: how do you measure results, and what exactly will the reports contain?

This is the first question to open any conversation with. A serious firm can describe specific indicators: presence rate across AI systems, comparison with competitors, the number of test queries and the models covered. An answer along the lines of "we will optimise your visibility in AI" with no specific metrics is a red flag.

What to look for: whether the provider can state how many test queries they use, which AI systems they check, how often they report and what exactly the report contains. They should also explain why they chose those metrics - that indicates understanding of the mechanism, not just selling a service.

Question 2: can you show a case study with measurable results?

Not a deck of general statements, but a specific case: a company in sector X, the starting point (presence rate Y%), the actions taken, the result after Z weeks (rate W%). An anonymised case study is acceptable, but it has to contain numbers. "The client was satisfied" is not a case study.

A bonus: ask for a case study from a sector close to yours. A provider with experience in your category understands the specifics of the questions your buyers ask, and selects the test question set accordingly. An agency that has only done AI Visibility for e-commerce may not understand the specifics of B2B professional services.

Question 3: what exactly does the service include - a list of actions?

AI Visibility is not one activity but several parallel strategies: site content optimisation, external mention-building, technical work (robots.txt, llms.txt, schemas) and monitoring. A serious provider can list the specific actions and tie each to an expected effect.

• Content optimisation - what specifically changes on the site, and why does that help AI Visibility?
• Mentions - on which platforms, in which formats, at what frequency?
• Technical - do they check and fix robots.txt, llms.txt and JSON-LD schemas?
• Monitoring - how often and how do they measure the effects?

Question 4: how long before I see the first results?

Realistic expectations are an important honesty signal. AI Visibility does not work in a week. Site content optimisation can improve citations in models with real-time access - Perplexity, Google AI - within two to four weeks. Building external mentions, and effects in models' training data, takes two to four months.

A red flag: promises of immediate results, or a guarantee of appearing in ChatGPT's top three. AI is non-deterministic, so no provider can guarantee a specific position in a model's answers. They can guarantee actions and systematic measurement - results follow from the sum of the work, not a magic switch.

Question 5: do you run a diagnosis first, or go straight to implementation?

Serious AI Visibility providers always start with a diagnosis, because without knowing the starting point you cannot plan the right work. A firm proposing implementation immediately, with no audit, is either offering a template instead of a tailored plan, or has hidden the diagnosis inside a higher price.

The ideal sequence: diagnosis, analysis of the results, an action plan, implementation, monitoring. You should be able to buy only the diagnosis and implement it yourself - a serious provider will allow that. If a firm sells only packages with no option to buy the diagnosis alone, it is worth asking why.

Question 6: how do you separate the effects of your work from natural market change?

This is a hard question that distinguishes reflective providers from those who take credit for every positive change. AI systems are updated regularly - a new model version can change results with no action on your part. A serious provider monitors not only your company but competitors and general model changes as well.

Good methodology means tracking the presence rate of your company and two or three competitors in parallel. If your rate rises while competitors' rates are stable or rising too, the effect is market-driven. If yours rises considerably faster, the work is producing an effect.

Question 7: what happens if the results are unsatisfactory?

An honest provider has an answer. It might be analysing the causes and correcting the plan, extending the period at no extra cost, or allowing termination if there are no results after a defined time. A firm with no policy for the absence of results transfers all the risk to the client.

A checklist before signing

• The provider can describe the specific metrics they will report
• They have a case study with measurable results, anonymised or named
• They list the specific actions included in the service
• They assess the time to results realistically - at least four to eight weeks
• They propose a diagnosis before implementation
• They have a methodology for separating their effects from market change
• They describe the terms for ending the engagement clearly
• They offer a work sample or a free preliminary diagnosis