How we measure AI search visibility.

Most AI search visibility tools and audits are opaque. They show you a score; they do not show you the prompts they tested, where those prompts came from, how brand mentions were parsed, or how sentiment was classified. Without that, the score is just a number to argue with.

We document everything below because (a) you deserve to judge our methodology before you trust our conclusions, and (b) the entire audit category is too new to have settled norms — making our approach public lets us be corrected when something is wrong.

The prompts we test must reflect how real buyers in your category actually phrase their questions — not keyword permutations a marketer guesses at. Generic prompt sets produce audits that look thorough but fail to represent buyer reality. Our prompt set construction uses proprietary sourcing methods spanning real buyer language, competitor-validated framings, role-conditioned variants, and category-specific pain phrasing.

Every prompt in every audit ships with its source category and credibility tier so you can judge the methodology in front of you. The exact mix and weighting evolves with each audit cycle as we learn what produces signal.

Different AI engines cite different sources, weight different signals, and surface different brands. A single-engine snapshot misrepresents reality. Our paid tiers cover all four major answer engines, with both programmatic and manually-captured runs depending on what each engine's API permits.

Free audits (oppority.com/check) provide a reliable indicator using one major engine. Paid tiers expand to cross-engine coverage where the operational cost is justified by the engagement.

Whether you appear in an AI answer is the floor, not the ceiling. Our scoring captures multiple dimensions that traditional "mention or no mention" tools miss:

• Position within the answer — AI tends to lead with the brand it most trusts; later positions matter less for buyer conversion
• Frequency across prompts — consistent recall across the prompt set is a stronger signal than a single hit
• Intent weighting — high-purchase-intent prompts (comparison, product-category) carry more weight than top-of-funnel discovery prompts

The composite score sits on a 0–100 scale that lets you see movement over time. The exact weighting and decay function are proprietary to our engine.

For every response, we identify the websites AI engines actually pulled from as evidence. Aggregated across the prompt set, these citation surfaces reveal the small number of sites that disproportionately drive what AI says about your category. Earning placement on those surfaces is the highest-leverage move available — and the one most companies overlook entirely.

We handle the technical details of resolving grounding redirects so you see the actual source domain, not a wrapper.

Being mentioned is not enough. How you are framed matters for buyer conversion. Each brand mention is classified along a tone spectrum that ranges from enthusiastic recommendation to dismissive caveat. The weighted sentiment score reveals not just whether AI recalls your brand but whether it positions you favorably relative to alternatives.

Every classification in every audit ships with a short rationale so you can audit our auditor.

Same product category, but framed by specific buyer roles. A generic "best compliance software" query and a role-conditioned "I am a Compliance Officer at a healthcare clinic looking for..." query often surface entirely different brands. Persona-conditioned testing reveals which buyer journeys map to your brand and which do not — a dimension most existing tools do not measure.

The persona set used in each audit is customised to your actual ICP, not a generic template.

Critique-elicitation prompts reveal whether your brand surfaces as a recommended alternative when buyers express dissatisfaction with competitors. This is one of the most valuable visibility opportunities most measurement tools do not capture — and one of the leverage points where a small amount of well-placed work can produce disproportionate movement.

The AI search visibility category is two years old. Open questions remain across the industry — not just for us:

• Stability of AI rankings across days and weeks. AI engines drift; the right way to model that drift is still being researched.
• Time-to-effect for visibility-improvement work. Schema changes, comparison content, and new citation placements show up in AI answers on different timescales (typically 4 to 12 weeks). Predicting this precisely is hard.
• Cross-engine consensus. Whether visibility in one engine predicts visibility in another is an open empirical question.
• Decay rates. How visibility scores degrade over 6–12 months without active maintenance is an area we are actively building data on.

We do not pretend to have these solved. We do measure them honestly and update our engine as we learn.