one-off · fixed price · fixed scope

AI Search & SEO Visibility Audit

Which channel to fund first, and the evidence behind the answer.

Your buyers now build their shortlist inside ChatGPT, Perplexity and Google's AI Overviews, before they ever reach your site. This audit measures what those assistants say about your category, maps the sources they read to say it, finds the search openings you already half-own, and ends with one recommendation you can act on: the channel to fund first, and why.

Book a scoping call $1,500 one-off · ten business days · two audits a month

Scope is settled before you pay, on a short call or by email, so you know exactly what you are buying. If the audit is the wrong purchase for you, we say so first.

what you decide at the end
One channel, named

Which of Reddit, X, LinkedIn or email to fund first for your category, with the evidence that led there and the reason the others wait.

A ranked list of fixes

The things to change, ordered by hand with the reasoning printed beside each one, and tagged client-side, one-off or retainer.

A 90-day expectation

What moving this actually looks like on your timeline, stated before you spend, including where it is slow.

What is in it

Six deliverables. The document is short on purpose: a long report is a way of not deciding.

01

A frozen buyer-question panel

20 questions in five groups (category, comparison, problem, direct brand, buying criteria), written the way your buyers talk to an assistant rather than the way keywords look. It is yours to keep and re-run for as long as the surfaces exist, whether or not you work with us.

02

Every answer, recorded and scored

The panel run across ChatGPT with search, Perplexity and Google's AI Overviews, in fresh sessions inside a single 48-hour window, with country and language pinned and stated. You get your mention rate, your first-mention rate and your share of all brand mentions against three to five named competitors, with the denominator printed beside every number. Questions where a surface returns no AI answer are recorded as no-answer and still counted.

03

The citation source map

Every domain and URL the assistants cited across the panel, ranked by frequency, tagged by type, with the specific pages and threads carrying each competitor's mentions. It is the shortest honest description of where your category's evidence actually lives, and it is the part that survives longest, because sources move more slowly than answers.

04

Site readability for assistants and crawlers

AI crawler access checked bot by bot, server-rendered versus JavaScript-only content, llms.txt presence and factual accuracy, schema coverage and validation, and whether your headings are shaped so an assistant can quote them. Covers sites up to 250 crawled pages; larger sites are quoted before you pay.

05

The search opportunity list

Four countable kinds of opening: keywords you already rank 4 to 20 for on pages that already exist, gaps where competitors rank and you do not, panel questions with no answer-shaped page behind them, and the community threads worth answering. Every volume figure is labelled a third-party estimate and carries its source and date.

06

A two-page decision memo, plus the raw data

One channel to fund first, the fixes ranked by hand with the reasoning printed next to each one, what to expect in 90 days, and what not to buy. Behind it: both answer tables as CSV, the archived text of every recorded answer, the frozen panel file and the itemised cost of the data used.

How it runs

  1. You send the intake. Your site, three to five competitors, who you sell to, the country and language to pin, and the objections you hear on sales calls. About 20 minutes of your time.
  2. The panel is built and frozen. Questions are mined from real search data, from the People Also Ask boxes in your category, and from the way buyers phrase things in the communities they use. Once frozen, questions are never edited in place, because the value of the second run is that the questions did not move.
  3. The run and the read. The panel goes across the assistants inside one window. Then the slower half: reading the cited sources, checking the site, pulling the search openings, and deciding what actually matters in what order.
  4. The memo. Delivered in writing within ten business days of the intake. A 30-minute walkthrough is included if you want it, and the document stands on its own if you do not.

What this audit will not claim

Printed here rather than buried in the report, because the limits are how you tell a measurement apart from a pitch.

No guaranteed citations and no guaranteed rankings. Nobody controls what an assistant answers, and a vendor who guarantees it is selling a coin flip with your money.
No claims about how assistants rank or select internally. The report describes citations observed on dated runs, not mechanisms nobody outside those companies can see.
Answers are dated, personalised and unstable. Every figure carries its run date, its surface and its sample size, so you can tell a real move from noise later.
Keyword volumes are third-party estimates, labelled as estimates, with their source and date.
No competitor data we cannot obtain honestly: no scraped private content, and no ad spend presented as fact.
No invented benchmarks. Where a number is ours rather than published, it says so and shows its sample size.
read the method before you buy it

The site-readability half of this audit has already been run on our own domain: the scores, the fixes applied and what is still open are on the proof page, unflattering parts included. Our own buyer-question panel is not published yet. When it is, it lands there too.

or run a version of it yourself

The measurement half is not a secret. Our 60-minute DIY procedure is free and complete, and if that is all you need, take it. This exists for teams who want the sources ranked, the site checked and the decision made.

Who buys this

· B2B teams who suspect they are invisible in AI answers and want the number before committing to a retainer
· Founders choosing between Reddit, LinkedIn, X and email who are tired of deciding on instinct
· Marketers who need an outside read they can put in front of a board or a client

Frequently asked questions

How is this different from an AI visibility dashboard?

A tool hands you a dashboard and leaves the interpretation to you, which is fine if interpretation is the part you already have. If a dashboard is all you need, buy the dashboard: it costs less than this and we will tell you so on the call. What this adds is the part a dashboard cannot: the cited sources ranked by whether you can realistically win them, the fixes ordered by hand with the reasoning printed next to each one, and a single channel recommendation. The panel is yours afterwards, so you can keep measuring without paying anyone.

How many of these have you run?

RouterGrowth launched in August 2026, so the honest answer is that this is a new product and you would be early. Rather than dress that up, the method is published instead of described. The site-readability half has already been run on our own domain, and the scores, the fixes and what is still open are on the proof page. The measurement half is written out step by step in our free DIY guide, which is the same procedure we run for you, with the tooling and the competitor scoring added. Being an early client of a new firm is a real trade: you get the founder's full attention and a method you can check line by line, and you do not get a decade of pattern recognition. Both halves of that are true.

Is the fee credited if we hire you afterwards?

Yes, against your first retainer invoice if you start within 90 days, and it is not a discount on the retainer. The frozen panel becomes the first month's AI share-of-voice report and the triaged thread list becomes the first month's target list, so crediting the fee stops you paying twice for the same artifacts. If the audit concludes RouterGrowth is not your lever, there is nothing to credit and the memo says so in writing.

What if the audit says we should not hire you?

Then it says that, in writing, with the reasoning. Every item in the plan is tagged client-side, one-off or retainer, and the client-side ones are not softened to make a retainer look necessary. Categories exist where the assistants read review sites and documentation rather than communities, and in those the honest recommendation is not our Reddit engine.

How long does it take, and what if it is late?

Ten business days from the intake. Two audits a month, so the work does not queue behind retainer clients. If the memo lands late, the fee is refunded and you keep the report.

Which assistants do you measure?

ChatGPT with search, Perplexity, and Google's AI Overviews, with the country and language pinned and stated in the report. Each question is run once per surface unless the report says otherwise, and the sample size is printed rather than implied.

Do we need this if our classic SEO is not fixed yet?

Do not sequence them, because the inputs overlap. Citable structure, clean pages and credible third-party mentions serve both, which is why the search opportunity list sits in the same document as the AI answers. If the audit finds that a technical blocker is the whole story, it says so on page one and the rest gets cheaper.

go deeper

$1,500, ten business days, yours either way.

Two audits a month, so the work never queues behind retainer clients. If the memo lands late, the fee is refunded and you keep the report. If you hire us afterwards, the fee comes off your first retainer invoice.

Book a scoping call See the ongoing service

Prefer email? Send your site and competitors to hello@routergrowth.com.