From buyer questions to your next useful move.

Start with three inputs

Provide your company website, market or country, and language. The intended onboarding proposes a company profile, products, category, audiences, use cases, competitors and buyer questions. You confirm or edit the suggestions; an inferred competitor is not accepted as fact without review.

No GA4, CRM, Search Console or CDN-log connection is required for the initial answer-measurement workflow. Referral and crawler reporting are separate, optional signals with their own setup.

Approve questions that matter to buyers

Start with approximately 20–25 questions across Discovery, Comparison and Decision. A discovery question asks which kinds of products solve a problem. A comparison question examines alternatives. A decision question tests concrete requirements, constraints or tradeoffs.

Not every question should name your brand. Use the language of a buyer who may not know you exist. Keep the original question and subsequent versions so a trend does not silently compare different inputs.

Run a baseline on a declared surface

The initial beta targets OpenAI and Perplexity API/search surfaces, using Mentionloom’s server-side credentials. Customers are not expected to purchase consumer AI subscriptions to supply API access.

Every report must identify the surface, model, market, language, collection window and actual number of samples. API output must not be presented as a direct recording of a consumer app. Coverage will be confirmed with the invitation; the website’s current previews are illustrative.

Open the evidence behind a losing question

Read the raw answer before acting on the score. Was your brand absent, simply mentioned, shortlisted or explicitly recommended? Which competitors qualified? Which sources were actually cited? What remains uncertain?

Recommendation Share and Lost Recommendation Share use the definitions in our methodology. A mention-rate preview is not the same metric. Referrals and crawler requests must never be combined with recommendations into an artificial score.

Choose a change you can justify

A useful opportunity connects a buyer question to a verifiable evidence gap. For example, a product may genuinely support a required integration while its public documentation leaves the supported plan, setup steps and limits unclear. The appropriate change is accurate documentation—not an unsupported claim that the AI will favor you afterward.

Record the hypothesis, responsible person, URL, proposed work and success criteria. Then record what was actually shipped and when. Read how to investigate competitor recommendations for a worked decision process.

Re-measure without confusing correlation with causation

Repeat the protocol, compare the same question set, and show the noise band and untreated comparison set where available. Record provider updates and other work that could have changed the answers. One positive cycle does not prove your edit caused the improvement.

The intended loop is baseline → opportunity → shipped improvement → re-measurement → observed outcome. Our measurement guide explains how to establish a baseline manually while evaluating this approach.

What is available today?

Mentionloom is in early access. Public Acme examples are fictional and are not live probes of your company. Provider coverage, sampling limits, pricing and integrations will be confirmed before access. No ranking, traffic or revenue increase is guaranteed.