A answrap.
Answer-engine visibility audit

Why does ChatGPT recommend them, not you?

Because it is quoting other people's pages — directories, review sites, roundups — and you are not on them. We ask your category's forty buying questions on ChatGPT, Perplexity and Claude, record which pages get quoted back, and hand you the list of the ones answering without you: which competitor is already there, exactly what it takes to get in, and the corrected files themselves.

ChatGPT Perplexity Claude
Asked through provider APIs, which are not the consumer apps. Perplexity is asked on its default fast preset — what a free account gets, not Pro, which is what most of your buyers are actually using.
Gemini: crawler access checked (Google-Extended), citations not sampled. Google AI Overviews has no crawler of its own — it is part of Search and follows your Googlebot rules. Every one of those choices is printed on the cover of your report, in the same size as the engine's name, because a setting nobody mentions is a claim nobody can check.
hello@answrap.com
One paid audit: €490 — €245 for the first five clients
free two-page snapshot in 24h
Send a domain and two competitors.
No logins, no analytics access, no admin accounts, ever.
report fragment illustrative
acme-crm.com
60 / 100 · grade C
weighted across four pillars
Crawler access 25%92
Machine-readable structure 25%15
Content extractability 20%72
Entity & citation footprint 30%62
CRITICALNo JSON-LD on any page — engines are guessing what the company sells.
HIGHCompetitors own every “alternatives to you” page in the citation set.
01 — Method

What we measure, and what we then go and do

the fourth one is where the money is
25%

Crawler access

Whether OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended and Googlebot can reach your pages at all. If you are blocked, nothing else matters — but on the ten UK sites we read on 29 July, none of them were, so expect this pillar to come back clean and the finding to be further down the page. Google-Extended covers Gemini apps, not AI Overviews — AI Overviews sits inside Search and follows your Googlebot rules, which is why Googlebot is on this list.

25%

Machine-readable structure

JSON-LD, Organization and sameAs entities, FAQPage and Product schema. Engines work out what you sell from structured data first and marketing prose second.

20%

Content extractability

Server rendering, heading hierarchy, question-form headings, comparison tables. Several AI fetchers do not run JavaScript and simply see an empty page.

30%HEAVIEST

Entity & citation footprint

Wikidata, Reddit, G2, Capterra and the third-party roundups engines quote for “best X” questions. Where most brands lose — and where almost nobody is looking.

02 — What the audit buys

Every finding comes with the job that fixes it

Anyone can hand you a list of problems. A list of forty problems in no order is not an audit, it is homework. Every line below arrives with an address, a first step, and an honest estimate of what it costs to do.

Public data only. No logins, no analytics,
no admin accounts — nothing to revoke afterwards.
We find

The directory, review and roundup pages that answer your category's questions with your competitor on them and you absent.

We do

Name each venue, the way in, and what it costs you in time — G2, Trustpilot, Clutch, Crunchbase and whichever came up for your category: which of them you can add yourself this afternoon, which need an editor pitched, which need nothing but a form. We do not create or claim accounts on your behalf, here or anywhere. You hold your own logins — that is the same sentence as "no admin accounts, ever", and it has to stay true in both directions.

We find

Which of fifteen named AI crawlers your robots.txt lets through, and which it quietly shuts out.

We do

Quote the lines that shut them out, and name the replacement agent by agent. Your developer edits one file. Two minutes of their time. We know the shape of a good answer because ours is one: answrap.com/robots.txt admits 21 agents by name — more than we audit for, and more than any of the ten sites we read on 29 July, including an SEO agency that scored 96 on this pillar without naming one. Open our file and yours side by side; it takes ten seconds and it is the only vendor claim on this page you can check before you email us.

We find

Whether a machine can tell what you sell: Organization, sameAs, FAQPage, breadcrumbs, sitemap, llms.txt.

We do

Name every missing field against the markup you already have — property by property, checked against your live JSON-LD rather than against a generic checklist, so your developer is reading a diff and not a tutorial. Subtypes count: if you have marked yourself up as a Dentist we do not tell you that you are missing Organization. Most checkers do.

We find

Your name, address and website written differently in every place they appear — which is how a model ends up unsure you are one company.

We do

List every place it differs, with the exact string it should read, and name which profiles are missing the link back to your site. Editing them is a few minutes each, and it is your account, not ours.

We find

Page by page, not just the home page — up to twelve pages from your sitemap, and the report says how many that was out of how many you list: does the text survive without JavaScript, is there one H1, is any heading shaped like a question a buyer asks.

We do

Hand over the exact headings to use, taken word for word from the questions you are absent from — so the heading matches the question a buyer actually typed, not one a copywriter guessed at. The body copy under them is yours to write; nobody outside your company should be putting words in your mouth.

Day five, in your inbox
→ The venues you are missing from, cheapest first
→ The exact lines to change in your robots.txt, quoted from your current one
→ The structured-data fields you are missing, named one by one against the markup you already have
→ One detailed page: every finding carries the address of the artefact it came from — the URL, the file, the line. Check any of them in ten seconds
→ One plain-language page you can forward to anyone
→ The remaining targets, sorted by what they cost to reach
Not included, so nobody is surprised on day five

Writing blog posts, comparison pages or FAQ copy. Link building. Internal-linking analysis — we read the pages your sitemap lists, up to 400 of them, but which page links to which is classic SEO, it is not what decides a citation, and we do not pretend to measure it. Anything requiring a login to your CMS. Posting on Reddit or Quora as you: that has to be a real person at your company, openly, and we will say so rather than fake it.

03 — What we refuse to promise
“best crm for small agencies” · asked 5×
run 1run 2run 3run 4run 5
Same question, same day, same wording. Solid bar = the brand cited first. This is the shape of the thing anyone promising a fixed ranking is pretending not to see. Illustrative.

We cannot make ChatGPT recommend you. Nobody can.

Answer engines are non-deterministic and personalised — two people asking the same question get different answers. So we measure what is real: whether you can be cited at all, and whether that is improving. Any vendor quoting a precise, stable “AI visibility percentage” is over-claiming.

And where we could not measure something, the report says so on the line itself — not in a footnote, not as a zero. A pillar we could not reach gets no score at all, and the reason is printed in its place. That rule is why the score you get is worth arguing with: every number in the document is marked as measured, derived, or unknown, and you can tell which is which without asking us.

no guaranteed rankings no invented percentages evidence against every finding no score for what we could not measure
04 — Why now

The rules of this category are not written yet

In search, the pages that answer your category's questions were settled a decade ago and are owned by somebody else. Here they are not settled. That is the whole opportunity, and it has a shelf life.

Measured on our own domain, 31 July 2026
26 / 40

buying questions where the engine had not settled on an answer — three identical asks, three different sets of pages quoted back.

Run #20: our own frozen 40-question corpus, asked three times each through Perplexity, 120 calls. One engine, one domain, one day — so read it as a sample of how unsettled this category is, not as a law of nature. It is the number we have, and we are telling you exactly how thin it is.

An unsettled question is a question where the position is still open. There is no established arbiter site for AI answers the way there is for search rankings — no directory everyone defers to, no scoreboard everyone quotes. The pages the engines reach for today were mostly written for other reasons and are winning by default. Being on them now costs a profile and a paragraph. It will not stay that cheap.

05 — Pricing

Fixed prices, published

Full audit

Five working days · we measure, tell you what to change, and build the part that needs no logins
€245 one-time, first five clients
€490 from the sixth client on. The first five pay half because we ask one thing in return: permission to publish the work — your name on it — as a named case study. We have no clients to point at yet, and buying that with a lower price is more honest than buying it with adjectives. There is one paid audit, not a ladder of them: the recommendations and the finished files are the same purchase.
Everything in Site check, plus
  • 40 category questions, asked on ChatGPT, Perplexity and Claude — 120 live answers, every one recorded and quoted back at you
  • Where the three engines disagree about you — reported as a finding, not averaged away
  • The pages answering your category without you — ranked by what it costs to get on them
  • Which of your competitors is already on each one
  • The questions you were never once cited on, listed word for word
And exactly what to do about it
  • Every venue named, with the link to its signup page, what you need in hand before you start, and how long it takes
  • The wording to use — taken from the exact questions you are absent from, so the profile answers them
  • The exact robots.txt lines to change, quoted from your current file — your developer edits one file
  • The structured-data fields you are missing, named one by one against what your markup already has
  • Every place your name, address or website is written differently, listed so you can make them match
And then we build it, in the same price
  • Your robots.txt, written out in full — the corrected file with the changed lines marked, not a description of them
  • The structured data as a finished block — the business type, address, coordinates and services you are missing, every field filled from what your own site and listings already say. Nothing invented.
  • The profile links block, assembled — every listing of yours we found, gathered into one place so the engines resolve you as one company instead of several
  • Your headings, written out — taken word for word from the questions you are absent from. The heading is ours, the answer underneath stays yours.
  • One free re-check after you publish — we confirm the files landed and are read correctly. Costs you nothing and takes us minutes.
  • Two pages: one detailed, one plain-language to forward
  • The raw engine answers this run was built from, kept and re-checkable — so any figure in the report can be traced back to the reply it came from
  • Five working days · still no logins to your site
  • We do not write the words on your pages. That is a claim about your business, and only you can say whether it is true.
  • We do not open accounts or claim listings for you. No logins, ever.
  • Files come to you for review. We never touch your live site, so we can never break it.
Order the audit — €245 →or email first, if you have questions
After paying, send your domain and two competitors to hello@answrap.com — that is the whole input.
Same day

Site check

Free
  • 15 AI crawlers checked by name
  • Structured data, sitemap and llms.txt read — up to twelve pages from your sitemap, not just the home page. The report states how many of your listed pages were read, and how many you list in total.
  • 3 of your real buying questions asked live, once each, on one engine we name — with the pages that were quoted back
  • Every finding names the file it came from
  • Same day, no call, no card
  • One engine only. Whether the engines disagree about you is a paid finding, because seeing it takes more than one
  • Nothing is built for you at this tier — the free check reads and reports, nothing more
Most sites come back technically clean, and we say so rather than invent a problem. The three questions are where it usually gets interesting.
Ask for a site check →
Every month, on repeat

Watch

€190 / month
Not on sale yet, and we will say why rather than take the money. Monitoring compares this month against the last one, so it produces nothing until there are two dated measurements of your category. We open it when your second one exists — not before.
  • The same 40 questions, asked the same way — so the months actually compare
  • Events, not percentages: a new source entered the answer set, a competitor appeared, a page stopped being quoted
  • The queue re-ordered every month as the category moves, so you always know the next thing worth doing
  • We tell you the day your own site update breaks the markup — that is how most of this quietly reverses
  • A month with nothing in it is reported as a quiet month, not padded
  • Cancel any time, runs to the end of the paid period
No button on purpose.
One document, all your clients in it

Agencies — priced by category, not by brand

€490 per category · €90 per extra brand in it
Same price as a direct client — we do not run a second price list for agencies. Your margin is the multiplier, not a discount: one category at €490 plus eight of your brands at €90 is €1,210, or €151 a client.
  • One category covers every client you have in it — the forty questions are the category's, not one brand's, so the run is bought once
  • A client-facing page you can forward as it is, or with your logo on it
  • A per-brand work order for each of your clients inside the category
  • We never contact your clients, ever
  • Each further category is a further €490, on the same terms
  • Every figure traceable to the raw engine answer behind it, if your client asks where it came from
06 — Objections

The questions you are about to ask

01

Is this just SEO?

No — and the difference decides everything. SEO competes for a link’s position. This decides whether a model can extract and cite a claim about you at all. A site can rank in Google and still be uncitable.

02

Do you need access to my site?

No. Public data only. Nothing to provision, nothing to revoke afterwards.

03

Can you guarantee results?

No. And anyone who does is selling you something.

04

How long does it take?

Five working days for the full audit. Twenty-four hours for the free snapshot.

05

What if nothing is wrong?

You get a page saying so, with the evidence. Most sites are technically fine. The gap is almost always in whose pages get quoted, not in your own.

06

Why is it €490 and not €3,000?

Because it is a defined piece of work with an end. Agencies charge $3,000–15,000 a month and do more than we do — content, links, PR, a monthly meeting. We measure your category, name every venue answering it without you, and hand the list over. One price, one delivery date, no account management to pay for.

07

What do I actually receive?

Two pages and a job list. One detailed page with the evidence behind every finding, one plain-language page you can forward, and an ordered list of what to change — including the exact robots.txt lines and the structured-data fields you are missing. Not a dashboard to learn.

08

Does the audit go out of date?

Yes — engines change what they crawl and cite. The structural fixes hold. The citation landscape moves. That is what monitoring is for; a one-time audit is still the right place to start.

One domain, two competitors, one email.

hello@answrap.com
No phone, no form, no chat.