Methodology

How InView Measures AI Visibility

How we turn real buyer questions into a consistent view of where your brand appears in AI search, who gets recommended instead and where you can improve.

AI answers can change between platforms and over time. InView measures patterns across a defined set of customer questions rather than relying on a single search.

10 buyer questions ChatGPT · Claude · Perplexity Fully automated About 2 minutes
The process: six stages
01UnderstandWe read your site, and you confirm the details and the market your customers are in.
02AskTen realistic buyer questions are built for your category.
03TestEach is put to ChatGPT, Claude and Perplexity as a fresh test.
04MeasureEvery response is recorded and judged in real time, with no human review.
05CompareYour position is set against every provider that appeared.
06RecommendUp to three actions point to what the evidence suggests is worth changing.

Independent tests

Each question is run independently, without conversational history from previous tests. Unbranded questions do not name the business being audited.

Competitive landscape

We show which businesses are recommended most often, how often each appears, and the types of question where the business appears or is missing, such as feature, category, use-case, comparison and problem-led questions.

Recommendation Rate
9 of 30

Does AI recommend the business? How many scorable answers recommended the business, shown by platform and by type of question. An answer that names no business, or couldn't be fully analysed, isn't counted.

Fictional example (Saltbush Sofas), illustrative only
Who AI recommends instead
#4 of 40

Who else gets recommended? The number of answers that recommended each business, counted once per answer, and the questions each business wins.

Fictional example (Saltbush Sofas), illustrative only
What AI says about you
2 of 3

Does AI know who the business is? Whether each platform describes the business correctly when asked by name and with more context. Quotes are shown word for word from answers about the business.

Fictional example (Saltbush Sofas), illustrative only
How to read your report

Four questions, answered from 30 AI answers.

  • Each answer is marked as recommended you, recommended others but not you, or didn't name any business. An answer that names no business is never treated as a failed recommendation.
  • The buyer market you chose is stated at the top. Most of the questions are written the way a buyer in that market would ask them.
  • The identity check shows what each platform said when asked about the business by name, and with more context.
  • Actions are limited to three. A suggestion is left out when the evidence doesn't support it, so some reports show fewer.
Methodology versions

Each report states how it was measured.

Each report is a snapshot of the defined tests at the stated time. Material methodology changes are versioned rather than silently changing results.

Current methodology: version 3 (November 2026). Each scan records the version it was measured under, and earlier reports keep the results they were measured with. Version 3 made four changes:

  • An answer that doesn't recommend any business, or that couldn't be fully analysed, is left out of the Recommendation Rate. Earlier versions counted these answers as not recommending you.
  • A product is counted under the business that makes it (for example, a named sofa under its brand) when the answer makes that ownership clear. Your own products count as you, never as competitors.
  • The written summary describes patterns in words. Every number in the report is calculated directly from the results.
  • You choose the buyer market the questions are set in. The report states that market.
What InView doesn't claim

The limits of this measurement, stated plainly.

01

We don't claim to know a secret AI ranking algorithm. AI platforms do not expose a complete explanation of why a provider was recommended.

02

We don't claim every customer asks the exact questions we test. The set is designed to be realistic and consistent, not exhaustive.

03

We don't claim results will never vary. AI outputs are probabilistic and change with model updates, sources, geography and timing.

04

We don't claim AI visibility replaces SEO. It is another layer of visibility alongside the work you already do.

05

We don't claim correlation proves causation. We distinguish between what we observed, what the evidence suggests, and what is worth testing.

Businesses named in reports reflect providers surfaced in the measured third-party AI outputs. Appearing more frequently does not mean InView has independently determined that provider is objectively better.

Payment never affects measurement. A commercial relationship with InView does not change how a provider is counted, scored or included, or which competitors appear in the underlying results.

Run the measurement on your own business.

An InView report is a structured snapshot of the defined tests at the stated time, not a permanent ranking or guarantee.

Check my AI visibility → Free · No card required · About 2 minutes