ES

Being mentioned is not being recommended.

When someone asks ChatGPT, Gemini or Perplexity for the best option in your category, the answer names a handful of brands. We measure whether yours is one of them — and then we work the sources that decided it.

Appearing in an answer is not the same as being the answer.

We commit to the study and the run. The recommendation, nobody controls.

Request the study

Measured across

What an answer looks like

The person asks

What is the best sunscreen for kids?

The assistant answers

For children, pediatric dermatologists usually recommend mineral filters. The brands mentioned most often are Bloqsol and Dermalux, both available in Mexican pharmacies, and the consumer agency guide has a comparison table.

Analysis, your brand

Solaria does not appear in the answer.

What influenced it

pharmacy retailer page · consumer guide · dermatology site

Solaria sells sunscreen in Mexican pharmacies. Its absence was decided by sources the brand does not control.

what's the best sunscreen for kids with sensitive skin?

ready

For sensitive young skin, dermatologists usually point to mineral filters. Bloqsol Kids is the one most often recommended for daily use, and Dermalux Pediatric comes up frequently for children with eczema. Look for zinc oxide, SPF 50 or higher, and a fragrance-free formula.

cited aad.orgconsumerreports.orgfarmaciasanpablo.com.mx
your brand competitor

Now click again. The same question, a second time.

Solaria did not appear at all. Two competitors did, and three sources were cited. On this ask, the brand is invisible.

Three answers, three shortlists. Solaria appears in one. Screenshot the first and the brand looks invisible. Screenshot the second and it looks fine. Both are a sample of one.

The full report page, on the invented sample: the consumer sample one-pager.

Google AI Overviews and AI Mode

Tracked in every visibility study

AI Overviews and AI Mode are not chat assistants. They are Google's results page writing the answer itself, above the links or in place of them. Because they always search, they show which pages Google trusts for a question, the same sources our studies track for every other assistant.

A written answer at the top of the normal results page. The links are still there, pushed down.

What a study records for each question

Both surfaces are included in the studies we run, reported apart from each other and from the chat assistants.

See how a study works

The problem, in the user's words

Users stopped searching. They ask. The sentence "ChatGPT doesn't recommend my company" usually arrives late, from a prospect who already chose a competitor.

The answer to a category question usually names three to five brands. The moment of loss is quiet: the competitor gets named, you do not. And the brands left out do not even find out they were left out. No report shows it. No alert fires.

You cannot fix what you do not measure.

What we do

We built the measurement tool and we run the agency around it. answer.show is the instrument. It is explicit about where it stops: it does not fix your site, commission your editorial content, or negotiate with your retailers. That is the agency.

01

Run the study end to end

We build the prompt bank, run it, read every response, and deliver the report.

02

Present the report

A working session with your team: what each rate means, and what it does not.

03

Train your team

To read the results and use them without us in the room.

04

Turn findings into priorities

What to fix first, and briefings for sales and marketing.

05

Work the sources

Retailer pages, commissioned editorial, earned media, and the site work the evidence justifies.

Then monitoring re-runs the same frozen bank, so you see whether the numbers moved, including when the honest answer is that they did not.

What gets counted

Four outcome rates, read the same way in every response of every run. Each one prints its denominator, because a percentage with no population behind it is a decoration.

26% 103 of 396 responses

Mentioned. The brand appears in an answer to a question that never named it. The floor: presence in the category conversation.

11% 44 of 396 responses

Recommended. The assistant actively suggests it. Being listed among eight is not being told to buy it.

3% 12 of 396 responses

Named as the best. The answer distinguishes it instead of grouping it. The strongest and rarest form of presence.

7% 28 of 396 responses

With a purchase path. The answer says where to buy it. The closest honest proxy to purchase intent, with no analytics needed.

Also counted: competitor mentions, every cited source, and refusals recorded as their own status. A refusal is information, not missing data.

How a study runs

  1. 01

    Kickoff research

    Your category, your buyers, and the decision the study must inform.

  2. 02

    Prompt bank

    Real buyer questions in three layers. Frozen with your approval.

  3. 03

    The run

    The same bank across assistants, repeated, so variance is visible.

  4. 04

    Reading

    Every response read and coded by a person, the same way each time.

Themes come from documented demand: Search Console queries, autocomplete, People Also Ask, forums, reviews. The full method is public: the method.

Where the lever is

What a model remembers from training is weak. The open web is what carries the brand. That is why the work after measurement is mostly editorial, retail content and PR, not a website migration, and why we do that work on the sources once the study names them.

The split by assistant type, why repetitions matter, and how a study is sized: the method.

Proof

Published method

Bank construction, models, repetitions, what gets counted, and the limits.

Read the method

Sample reports

The real deliverable, on an invented brand, for both segments.

Consumer B2B

Open data

Every response saved and readable. Every rate printed with its denominator.

What we believe, and the limits

We agree with the skeptics

Google: "optimizing for generative AI search is still SEO". The same guide calls llms.txt, chunking and special schema myths. We sell none of them.

Search Central OpenAI on OAI-SearchBot

Request the study

What to unlearn

  1. There is no position one in AI answers.

    Whoever sells you one invented it.

  2. Most of the answer was written by someone else.

    Third-party sources, not your site.

  3. The same question answers differently each time.

    One screenshot is an anecdote.

  4. A topic can matter before anyone searches for it.

  5. The first run is a baseline, not a fix.

Full versions and what each one replaces: /metodologia/.

Who it is for

Brands.

Consumer brands. Your buyer walks: assistant question, retailer app, reviews, shelf. Visits to the brand's own site on that walk: zero. The sources that decide the answer sit outside your control, and the study tells you which ones move it.

B2B and SaaS. Buying committees ask assistants early, in evaluation questions you never see. The study shows who gets named instead of you, and which sources put them there.

Start with the audit Then, monitoring

Agencies.

Offer AI visibility without building tooling: wholesale runs and white-label reports. The 360 program (tool, methodology, training, go-to-market) is the full package for agencies running AI visibility as a service line.

Talk to us about partners

Bring the decision, not the brief

The right first step is not software and not a package. It is the question your business needs to answer: does the category conversation include us, and who is in it instead?

Request the study. It starts with a conversation about that decision, not with a demo.

Request the study