Second Dove Let’s audit your business

Methodology

How we measure visibility across ChatGPT, Claude, Gemini, Google AI Overviews, and Perplexity: the five surfaces customers use when they ask an AI for a recommendation.

Scope Today’s question set was built for medspas. Other local service businesses are handled case by case: ask us.

Measure

We ask five AI surfaces the questions your customers ask, and count how often each one names your business.

These are the five surfaces where a customer, today, can ask a question and get a direct recommendation that names specific businesses. Perplexity became the fifth in June 2026, once it was a primary answer engine for local discovery.

Measured against ChatGPT’s default model with web tools enabled.

The question set

Every audit asks 36 types of question about a business. They fall into seven kinds. Four surfaces answer all 36 types; Google AI Overviews answers only the cost and consideration ones, half of the set.

  • DefinitionVisibility is the share of customer questions where an AI surface names your business in its answer.
  • Counted onceQuestions are the unit, not question-and-surface combinations, so the same question is never double-counted across surfaces.

One answer, three records

Mention
Named in the answer
Citation
Its website cited as a source
Rank position
Named first among competitors
Question:Example

best kitchen remodeler near me?

The question
we test

For a kitchen remodel nearby, many people start with Example Remodeling Co., then Sample Builders.

Invented names, shown to explain what we record.

Named in the answer

Slide to reveal
   the source

Questions customers type

Example

Invented examples, rotated across five trades to show the shape of each kind of question. The question set itself is written for medspas.

Service-led
who does brake repair near meExample, auto repair
Location-led
tattoo studio in SunnyvaleExample, tattoo studios
Comparative
[your barbershop] vs [rival]Example, barbershops
Trust-led
most trusted kitchen remodeler in SunnyvaleExample, contractors
Cost-led
how much does botox cost in SunnyvaleExample, medspas
Research-led
how to choose an auto repair shopExample, auto repair
Fit-led
tattoo artist taking walk-ins this weekendExample, tattoo studios
Technical detailWe call the full set the query basket.

The medspa set has 36 patient query types, expanded into 162 query-runner combinations: the 36 questions run across the four chat surfaces (144) plus 18 cost and consideration questions routed only to Google AI Overviews.

Four surfaces answer all 36 question types. Google AI Overviews answers only the cost and consideration ones.

Which surfaces are asked which kinds of question
SurfaceCost and consideration questionsAll other questions
ChatGPT, Claude, Gemini and PerplexityAskedAsked
Google AI OverviewsAskedNot asked

The 162 combinations and the visibility rate in your report measure different things, so they never tally directly. Your report’s rate is drawn from a smaller, service-specific figure: only the questions relevant to services your own site shows you actually offer count toward it (your “offered” basket).

Questions about services we don’t find advertised are tracked separately as market demand, never counted as a visibility gap. The unit is unique questions, not query-and-surface combinations, so the same question is never double-counted across surfaces.

Cost and consideration questions run per service, not just in aggregate. A business can be named for the “best” questions and still be missing where a customer checks the price or weighs whether it is worth it, which is where the booking is decided.

Each of the 36 queries is run against ChatGPT, Claude, Gemini, and Perplexity, and the 18 cost and consideration queries are also run against Google AI Overviews. We vary the phrasing slightly, because the same intent asked three ways often returns three different sets of businesses.

What we record for each answer

For each question on each surface we record three things: whether the business was named, whether its URL was cited as a source, and where it ranked among the named competitors.

  1. Before counting, we collapse variant spellings and sub-locations of the same chain into one brand.
  2. We do not collapse separately named businesses under the same owner.
  3. The per-surface visibility score is a 0 to 100 number reflecting mention rate weighted by rank position, normalized so that showing up first in every query equals 100 and showing up never equals 0.

02 Improve

The audit, on paper.

The method ends in a document. These are all three pages of our public sample report.

You get a short, ranked list of fixes, not a 40-item checklist. Each audit hands you one visibility index and the five changes that move it, ordered by projected lift. Your report ranks the fixes for your own gaps.

A finished
  report example

Page 1 of the sample report: one large percentage, how many AI answers named the business, its rank among the businesses AI names, and three industry statistics with their sources.
Page 2 of the sample report: a visibility score out of 100, and a bar chart of how many times each business was named, with the sample business among anonymized competitors.
Page 3 of the sample report: for each of the five AI surfaces, how many answers named the business and who is named most, then where the AI sources its answers.

Our public sample report, August 2026, issued under our former name, Sunnyvale AEO. The business and its competitors are anonymized.

Open the full PDF

Delivery Within 24 hours of intake, we walk the PDF to your front desk.

Prove

After the first fixes ship, we ask the same questions again and show you what moved.

If you go ahead, the next step is a 30-day re-measure: we ship the first fixes, re-run the same audit, and show you what moved. For active clients we re-score every month.

A re-measure, one real client · 3 of 5 surfaces

See what a re-measure looks like, and what it does not prove.

19%from 6%

  • Sep 10 · 6%
  • Oct 12 · 19%
of answers from ChatGPT, Gemini and Perplexity named one client, the same questions a month apart. Uses three of the five surfaces.

Read the case studySee the figure and its limits

Limits

Four things we do not claim.

  • AI outputs vary.We run multiple combinations per query type to smooth this, but two audits a week apart can disagree at the margin on lower-scoring brands.
  • The audit is a point-in-time snapshot.AI surfaces update continuously; we re-score monthly for active clients.
  • No position-one guarantees.We optimize for the underlying signals that AI surfaces weight, not for any single query result.

The 30% commitment has a definition.

It counts the unique customer questions where at least one of the five surfaces names you, measured at the 90-day mark against the baseline audit on the same query basket (unique questions as the unit, not question-and-surface combinations). Where your baseline names you in fewer than 5 unique questions (including zero), it is instead an absolute increase of at least 3 additional unique questions where an AI surface names you, since a percentage of so small a base is not a meaningful measure.

It is also conditional on your cooperation: it applies only if you grant and keep the access we need, respond to our requests on time, and stay through your third monthly payment or the 90-day mark, whichever comes first, with the 90-day measurement window extended for any client-caused delay. The full conditions are in our Terms of Service.

Questions about the method?

We’re happy to walk through any of this in detail before you commit. Email [email protected] or request your audit.

Let’s audit your business