Own-site research · study in progress

Answari’s AI visibility study.

Evidence status

Incomplete study. This dated record contains measured answers, unresolved checks and planned runs. It is not a finished baseline or proof of improvement. Missing or unreviewed answers stay unknown.

We are applying our measurement method to Answari itself. The questions cover finding a provider, understanding AI visibility and checking who is behind the business. This record shows what was captured, what could be assessed and where an answer gets the details wrong.

Some answers name Answari and cite our website. That does not make every description accurate. The useful question is what a prospective client would actually learn, and whether the supporting link and the facts hold up.

How much of the planned study is complete?

Planned observations
96
Answers recorded
68
Both checks assessed
64

4 partly assessed · 0 recorded and awaiting review · 28 wholly unassessed, including 2 saved access-gate attempts and 26 cells without attempt evidence. An observation is one exact question, in one product, on one run. These are study-completion counts, not visibility scores.

Swipe or scroll to see all columns.

Observed presence and coverage. Each presence count uses its own assessed denominator. On a small, selected and incomplete panel, raw counts are more useful than a single percentage.
Question groupName present / assessedMention coverage / plannedOwned link / assessedCitation coverage / planned
Non-branded questions4 / 5656 / 804 / 5252 / 80
Branded questions11 / 1212 / 168 / 1212 / 16

All four non-branded positives came from the same Plymouth or Devon question, A9. Three were ChatGPT repeats and one was a Perplexity answer. They are four observed responses, not four different buyer questions. See the repeat results and their limits.

A branded-name mention is not necessarily company recognition. Some branded answers use the name while discussing a surname or a different business. We retain the literal name result and flag the identity error. We do not call that a correct description of Answari.

View coverage by product

Swipe or scroll to see all columns.

Each product has 24 planned observations, including repeats.
ProductRecordedFully assessedPartial or pending reviewWholly unassessed
ChatGPT242310
Gemini242310
Claude10123
Perplexity191815

What do the recorded answers show?

These examples are observations from particular answers, not a claim about what every user sees. The register identifies their question, product, run and recorded time.

A9 · ChatGPT and Perplexity · run 1

An owned citation can sit beside inaccurate information.

For the non-branded question “AI visibility consultancy in Plymouth or Devon”, the captured ChatGPT and Perplexity answers both named Answari and linked to our homepage. Both were recorded at around 13:41 UTC on 5 September.

The Perplexity answer repeated older fees and a broader product scope. ChatGPT also used a broader product description. Our own published material still contained broader scope wording during capture; we corrected that afterwards. The link was real, but it did not guarantee current details. Our current services and pricing are the reference for the offer.

These observations do not establish when each product retrieved a version of our material, what caused its description, a ranking, uplift from site changes or a client enquiry.

A9 · the same local question, repeated

All four non-branded positives come from one question.

Across the assessed non-branded panel, all four name mentions and all four owned-site citations came from A9: “AI visibility consultancy in Plymouth or Devon”. They were repeat answers to one question, not four different buyer questions.

Swipe or scroll to see all columns.

Observed A9 answers only. Fractions use assessed answers, not all planned runs.
ProductName / assessedOwned link / assessedAnswers / planned
ChatGPT3 / 33 / 33 / 3
Gemini0 / 30 / 33 / 3
ClaudeNot assessedNot assessed0 / 3
Perplexity1 / 21 / 22 / 3

ChatGPT named and cited Answari in all three captured A9 runs. Gemini did so in none of its three. Perplexity did so in one of two; its third planned run remains unassessed after capture stopped at sign-up gates on other questions. No A9 answer was captured from Claude.

The results varied by product and, for Perplexity, between repeat answers. These are few selected responses under differing conditions. They do not establish population rates, a product-wide ranking or uplift caused by site changes.

B2 · Gemini · run 1

A name mention can identify the wrong business.

When asked who founded Answari, the captured answer described a different kind of business and gave an incorrect founder. It did not cite our website. The correct founder is Stacey Smith.

That answer contains the name, so the literal mention is recorded. Its identity error is kept alongside it. Treating the mention alone as success would hide the more important finding.

Source inspection and question quality

Unknown and off-target answers need their own labels.

Some source groups or destinations could not be fully resolved. Those citation checks remain unknown. Other answers interpreted a question differently, including a short GEO question answered in terms of geology. We retain those responses and flag the mismatch.

An unresolved source is not evidence of no citation. An answer to a different question is not strong evidence about whether a buyer would find the consultancy.

How was this study designed?

The frozen panel contains 14 exact questions across ChatGPT, Gemini, Claude and Perplexity. Ten questions do not name Answari; four do. Each question has one planned run in each product. Five chosen non-branded priorities have two further runs in every product, giving 96 planned observations: 80 non-branded and 16 branded.

The five priorities are A1, A2, A5, A7 and A9. They were chosen for this own-site study, covering four national buyer or source questions and one Plymouth or Devon question. This is our research design, not a mandatory scope for every client.

Read the exact question panel

Swipe or scroll to see all columns.

Question wording is retained even where an answer exposes ambiguity. A revised question would belong to a separately identified panel.
IDExact questionPlanned runs
A1How do I get my accountancy firm recommended by ChatGPT and AI search?3 per product
A2Who can help my law firm appear in Google AI Overviews and answer engines?3 per product
A3Best generative engine optimisation (GEO) consultant in the UK1 per product
A4How can a financial advice firm show up when people ask ChatGPT for a recommendation?1 per product
A5AI visibility consultant for professional services firms UK3 per product
A6Who does answer engine optimisation for regulated firms (SRA or FCA) in the UK?1 per product
A7How do I make sure AI engines cite my professional practice as a source?3 per product
A8Best for GEO in Plymouth1 per product
A9AI visibility consultancy in Plymouth or Devon3 per product
A10GEO or AEO agency in Devon1 per product
B1What is Answari?1 per product
B2Who founded Answari / who is behind it?1 per product
B3Is Answari a good GEO or AI visibility consultancy?1 per product
B4What does Answari do and who is it for?1 per product

Actual product conditions are part of the evidence.

Capture used the normal consumer interfaces, with fresh conversations and temporary or incognito modes where available. Some products were signed in and others signed out. Exact model versions were not always displayed. The downloadable register records the observed mode and session limits for each captured answer.

Location and other personalisation were not independently controlled. A fresh conversation does not prove a context-free answer. Captures were interleaved across products and website releases. This is not a controlled before-and-after experiment.

The work used AI-assisted browser capture and saved-evidence review. Some capture and review were performed by the same AI assistant. This is an authorised own-site research record, not independent human validation.

Two checks, with separate denominators.

  • Name mentioned: does the answer contain the Answari name? Branded answers also need an identity check, because the name may refer to a different business.
  • Owned site cited: does at least one inspected source URL point to answari.co.uk? Several owned links in one answer still count as one cited observation.
  • Unknown: not run, awaiting assessment or not sufficiently inspected. Unknown observations are excluded from that metric’s assessed denominator and remain visible in coverage.

Recommendations, prominence, relevance and factual accuracy are additional observations. A third-party publisher’s link is not an owned-site citation. Read the full methodology and our measurement guide for the definitions and the distinction from native platform reports.

Inspect the observation register.

The downloads distinguish actual answers, access-gate attempts and cells without attempt evidence across all 96 planned cells, the exact questions, assessed flags, owned-link destinations, recorded times and public session metadata. They exclude private account identifiers and complete third-party AI answers. This is an observation register, not a release of the full raw evidence archive.

Snapshot prepared 2026-09-05T14:06:01Z. This dated snapshot remains incomplete; missing and unresolved observations are explicitly retained.

View the 68 recorded observations

Swipe or scroll to see all columns.

Yes and No are assessed results. Unknown is a gap. Public notes identify selected accuracy and question-quality issues.
ObservationRecorded time, UTCNameOwned citationStatus and note
A1-ChatGPT-r12026-09-05T12:51ZNoNoAssessed
A1-ChatGPT-r22026-09-05T13:51:36.767ZNoNoAssessed
A1-ChatGPT-r32026-09-05T13:57:14.162ZNoNoAssessed
A1-Gemini-r12026-09-05T12:59:18ZNoNoAssessed
A1-Gemini-r22026-09-05T13:50:47.272ZNoNoAssessed
A1-Gemini-r32026-09-05T13:56:09.653ZNoNoAssessed
A1-Claude-r12026-09-05T13:06:51ZNoUnknownPartly assessed
Source destinations are not completely resolved. Citation stays unknown.
A1-Perplexity-r12026-09-05T12:59:57ZNoNoAssessed
A1-Perplexity-r22026-09-05T13:52:30.710ZNoNoAssessed
A2-ChatGPT-r12026-09-05T12:53:27+00:00NoNoAssessed
A2-ChatGPT-r22026-09-05T13:52:28.531ZNoNoAssessed
A2-ChatGPT-r32026-09-05T13:58:11.627ZNoNoAssessed
A2-Gemini-r12026-09-05T13:29:59.192ZNoNoAssessed
A2-Gemini-r22026-09-05T13:51:40.296ZNoNoAssessed
A2-Gemini-r32026-09-05T13:57:17.737ZNoNoAssessed
A2-Perplexity-r12026-09-05T13:02:09+00:00NoNoAssessed
A2-Perplexity-r22026-09-05T13:53:58.078ZNoNoAssessed
A3-ChatGPT-r12026-09-05T12:55:35+00:00NoNoAssessed
A3-Gemini-r12026-09-05T13:32:13.759ZNoNoAssessed
A3-Perplexity-r12026-09-05T13:30:17.959ZNoNoAssessed
A4-ChatGPT-r12026-09-05T13:28:39.933ZNoNoAssessed
A4-Gemini-r12026-09-05T13:34:24.400ZNoUnknownPartly assessed
Source destinations are not completely resolved. Citation stays unknown.
A4-Perplexity-r12026-09-05T13:31:12.908ZNoNoAssessed
A5-ChatGPT-r12026-09-05T13:31:34.794ZNoUnknownPartly assessed
Source destinations are not completely resolved. Citation stays unknown.
A5-ChatGPT-r22026-09-05T13:53:57.419ZNoNoAssessed
A5-ChatGPT-r32026-09-05T13:59:06.177ZNoNoAssessed
A5-Gemini-r12026-09-05T13:36:53.065ZNoNoAssessed
A5-Gemini-r22026-09-05T13:52:30.365ZNoNoAssessed
A5-Gemini-r32026-09-05T13:58:12.561ZNoNoAssessed
A5-Perplexity-r12026-09-05T10:31:49ZNoNoAssessed
A5-Perplexity-r22026-09-05T13:55:05.652ZNoNoAssessed
A6-ChatGPT-r12026-09-05T13:32:55.183ZNoNoAssessed
A6-Gemini-r12026-09-05T13:39:56.578ZNoNoAssessed
A6-Perplexity-r12026-09-05T13:35:01.317ZNoNoAssessed
Off-target answer about authorisation rather than the requested suppliers.
A7-ChatGPT-r12026-09-05T13:35:01.006ZNoNoAssessed
A7-ChatGPT-r22026-09-05T13:55:03.462ZNoNoAssessed
A7-ChatGPT-r32026-09-05T13:59:41.296ZNoNoAssessed
A7-Gemini-r12026-09-05T13:40:32.113ZNoNoAssessed
A7-Gemini-r22026-09-05T13:53:57.740ZNoNoAssessed
A7-Gemini-r32026-09-05T13:59:10.304ZNoNoAssessed
A7-Perplexity-r12026-09-05T13:36:53.369ZNoNoAssessed
Off-target answer about citing AI in writing rather than the requested business question.
A7-Perplexity-r22026-09-05T13:56:10.001ZNoNoAssessed
Off-target answer about citing AI in professional work rather than earning a citation from AI.
A8-ChatGPT-r12026-09-05T13:36:28.562ZNoNoAssessed
The short prompt produces a geology/place interpretation and a clarification request.
A8-Gemini-r12026-09-05T13:41:46.985ZNoNoAssessed
The short GEO prompt is interpreted as geology and ground engineering.
A8-Perplexity-r12026-09-05T13:39:56.888ZNoNoAssessed
A9-ChatGPT-r12026-09-05T13:41:02.504ZYesYesAssessed
Names Answari and links to its homepage. Product scope differs from the current offer; our own published material also had broader wording during capture.
A9-ChatGPT-r22026-09-05T13:56:03.188ZYesYesAssessed
A9-ChatGPT-r32026-09-05T14:00:21.618ZYesYesAssessed
A9-Gemini-r12026-09-05T13:42:43.639ZNoNoAssessed
A9-Gemini-r22026-09-05T13:55:05.303ZNoNoAssessed
A9-Gemini-r32026-09-05T13:59:42.601ZNoNoAssessed
A9-Perplexity-r12026-09-05T13:41:02.203ZYesYesAssessed
Names Answari and links to its homepage. Repeats older fees and broader product scope. Our own material still contained broader scope wording during capture; it was corrected afterwards.
A9-Perplexity-r22026-09-05T13:57:18.092ZNoNoAssessed
A10-ChatGPT-r12026-09-05T13:43:02.666ZNoNoAssessed
A10-Gemini-r12026-09-05T13:43:35.142ZNoNoAssessed
A10-Perplexity-r12026-09-05T13:44:40.746ZNoUnknownPartly assessed
Source destinations are not completely resolved. Citation stays unknown.
B1-ChatGPT-r12026-09-05T13:44:39.754ZYesNoAssessed
Literal name appears, but the answer discusses a surname and does not recognise the actual consultancy.
B1-Gemini-r12026-09-05T13:05:38ZYesYesAssessed
B1-Perplexity-r12026-09-05T13:48:53.583ZYesYesAssessed
Fees match the current service page in this answer; the description blurs direct product tests with Google AI reporting.
B2-ChatGPT-r12026-09-05T13:48:50.668ZYesYesAssessed
Names the correct founder but describes a broader product scope than the current four direct products.
B2-Gemini-r12026-09-05T13:44:40.396ZYesNoAssessed
Literal name appears, but the company and founder are misidentified. The actual founder is Stacey Smith.
B2-Perplexity-r12026-09-05T13:49:57.686ZYesYesAssessed
B3-ChatGPT-r12026-09-05T13:49:55.465ZYesYesAssessed
Uses an outdated product scope. Positive wording is an AI opinion, not evidence of client results.
B3-Gemini-r12026-09-05T13:48:53.228ZYesYesAssessed
Positive fit opinion is an observed AI judgement, not a client testimonial or commercial-outcome proof.
B3-Perplexity-r12026-09-05T13:50:47.601ZYesYesAssessed
Distinguishes the company’s educational content from independently documented client outcomes. This is an observed opinion, not an exhaustive public-evidence search.
B4-ChatGPT-r12026-09-05T13:50:41.666ZYesNoAssessed
Repeats the name but does not identify the Plymouth consultancy. Describes unrelated similarly named businesses.
B4-Gemini-r12026-09-05T13:49:57.342ZNoNoAssessed
Answers about a different, similarly named service. This does not describe Answari.
B4-Perplexity-r12026-09-05T13:51:40.645ZYesYesAssessed
The retrieved services source-card snippet retains outdated fees. Those fees are in the snippet, not the generated answer prose. The answer also has product-scope ambiguity.

What can this evidence support?

It supports a dated account of the answers and source links we inspected under the recorded conditions. It identifies practical questions to resolve, including outdated descriptions, identity confusion and missing source evidence.

It cannot establish Answari’s share of all AI answers, prove that site changes caused a citation, or show a commercial result. This is a small, selected panel with incomplete and uneven coverage. Repeat responses are not statistically independent samples.

A citation is not a website visit. A website visit is not automatically an enquiry, and an enquiry is not revenue. Search Console, Bing Webmaster Tools, analytics and actual enquiries must be assessed separately, with their own dates and definitions.

Use the same discipline on your own business.

Choose a small set of real buyer questions, retain the exact wording and record what the answer and its links actually show. Leave missing checks blank. Verify the identity and details before calling a mention useful.

Our free AI visibility scorecard provides a practical starting point. For a scoped review of your practice, see the Visibility Audit.