AI Search Audit

Can AI find your business?

I check how clearly your website represents your business to AI search —
and show you what to improve, so the answer about you is the one you would give.

  • What do you do? clear
  • Who is it for? vague
  • Where do you operate? missing
  • What makes you different? vague
  • content
  • structure
  • entities
  • schema
  • crawlability
AI answer

A company specialising in [a service] for [a customer group], operating in [a region]. It stands out for [a trait].

A model does not guess better than the website tells it. The blanks stay blank — or a competitor fills them.
ChatGPT · Google AI · Perplexity · GeminiFixed scope, fixed priceSolo, 15+ years
The problem

People ask before they call

More people skip the ten blue links and ask ChatGPT, Perplexity or Google AI a question instead — who does this, who is nearby, what does it cost, who should I consider — and get one answer. If you sell a service people compare before buying — a hotel, a workshop, a clinic, an office, a manufacturer, a consultancy — that question about you is already being asked. You just never see the answer.

Most company websites were written for someone who already knows what the company does: a referral, a partner, someone who typed the company name. An AI system reads the same page like a stranger with one question and three seconds — what do you do, for whom, where, and why you. If the page does not say it plainly, the model guesses. It guesses in generalities, or it picks the company whose page said it plainly.

I am not quoting traffic shares or market percentages here, because I do not have a source I would stand behind. One question is enough, and the audit answers it concretely: when an AI system describes your company, does your website give it enough clear information to get it right?

The approach

I don't optimise for one magic prompt. I make the website easier to understand

AI visibility is not a meta tag or a trick in the copy. It starts with a website that says plainly what the business is (entities), answers the questions customers actually ask (content), lays that out in a readable structure (headings, links, page types), and can be read by a machine at all (access, structured data, plain-text versions).

It helps to know how these systems work: most of them do not answer from memory. Before writing a sentence about your company they search for and fetch pages, then build the answer from what they managed to fetch and understand — a mechanism called RAG (retrieval-augmented generation). So what counts is not whether the model has "heard of" the brand, but whether your website can be found, fetched and cited as a source. It is also why nobody can honestly promise a specific answer: what gets fetched is decided by the search layer underneath, and that layer changes.

Those are the same things that make a site understandable to Google, to a screen reader and to a person landing on it for the first time. So the audit does not end with "add a file and wait" — it ends with a list of content and code changes that make sense whichever model is fashionable next year.

The goal is not a website that sounds like AI. The goal is a business that is easy to understand.

Scope

What I check

Six areas, each with a concrete list of questions. The query set and the systems are written down in the report — not "the whole internet".

Brand and entity clarity

Can a machine clearly tell what your company is, what it offers and where it operates — without inferring it from context?

Content

Are the questions customers ask answered directly, clearly and in enough depth — an answer, not a slogan?

Structure

Are pages, headings, internal links and content types organised around real topics and entities, not around the menu?

Structured data

Does the site expose useful machine-readable information where it makes sense — organisation, services, FAQ, location?

Technical access

Can crawlers discover, fetch and understand the content they need: robots, sitemaps, canonical URLs, rendering, blocks?

AI visibility

How the brand appears in answers to a defined query set across chosen systems — what they say, what they omit, who they name instead of you.

The query set

I test the questions your customers actually ask

There is no single AI ranking — there is an answer to a specific question. So the first thing I agree with you is the set of questions you want to be the answer to, and that set is the measure before and after.

  1. Who builds [service] for [industry] in Poland?
  2. What is the best [service] for a company like [brand]?
  3. How much does [service] cost?
  4. What should a company look for when choosing [service]?
  5. Which companies offer [service] in [city / region]?
  6. What makes [brand] different?
  7. Does [brand] handle [segment / use case]?

These are templates, not results. The actual set — usually 15–25 questions — is written for your industry and region, and the report lists it in full, together with the systems I checked it in.

What you get

A list of things worth fixing

Not a generic 40-page SEO report — specific changes to your website, in the order worth making them.

Visibility snapshot

What AI systems currently understand about your business: what they say, where they are wrong (old prices, hours, address, a retired offer), where they go quiet, who they name instead of you.

Query set

The written list of customer questions and systems the audit was run on — so it can be repeated in six months and compared.

Findings

Concrete problems in content, structure, technical accessibility and entity clarity — each with its place on the site and an example.

Priorities

What to fix first, what can wait and what is not worth touching. No "everything is important".

Recommendations

Specific website changes: which sentence to add, which page to split, what to unblock for crawlers. Where the change is code — schema, llms.txt, robots — you get the snippet ready to paste.

Implementation, if you want it

I can also make the changes — in the content and in the code — and, when the foundation is the problem, move the site to a simpler stack.

Technically

A website that AI can read is usually a better website for everyone

Clear content structures, semantic HTML, useful structured data, accessible pages and machine-readable versions are not AI-only features. The same things help search engines, assistive technologies and a person on a phone. That is why the audit's list rarely contains anything done for the models alone.

One caveat: an llms.txt file and Markdown mirrors are one asset in the architecture, not a visibility guarantee. The standard is new and not every model honours it. I ship them where they make sense and say plainly what they do and do not do.

  • semantic HTML
  • clear headings
  • entity clarity
  • structured data
  • internal linking
  • crawlability
  • sitemap
  • canonical URLs
  • machine-readable content
  • Markdown mirrors
  • llms.txt where appropriate
Evidence

I already build websites this way

The audit is based on the same principles I use when building websites: structured content, semantic markup, machine-readable pages, technical SEO and a frontend that stays out of the way. I do not need a slide deck to prove it — the page you are reading is built exactly this way, and you can check it in a minute.

It also means the audit can lead directly into implementation when the recommendations require code rather than another report.

  • Check it on this page

    Append .md to this page's address and see what a model gets: the text alone, no navigation, no scripts. Every page on this site has one, and the build fails when one is missing.

    ai-search-audit.md
  • An index for models

    A separate llms.txt in English and in Polish, generated from the site content rather than written by hand. The coverage check is part of the build.

    /llms.txt
  • I described the method publicly

    A practical guide to AEO: llms.txt, structured data, content in answer format, measurement. No promises, concrete steps.

    AEO — how to optimise a website for AI
  • Ask AI about this site

    Google AI and Perplexity describe spoko.space the way it describes itself — who, what, for whom, where, what sets it apart — because it says so plainly. One query in Perplexity shows it live.

    Ask Perplexity
  • Shipped on a large site

    A parts catalogue with thousands of pages: structured data, llms.txt, machine-readable versions. Search engines are its main source of traffic.

    VW Polo 6R parts catalogue
What next

An audit is useful. Fixing the website is better

Three paths after the report. You choose after reading it, not before.

Path 01

I give you the roadmap

You take the findings and implement them yourself or with your own developer. The report is written so that works without me.

Path 03

I rebuild the foundation

When the CMS or the architecture is the problem, I move the website to a simpler stack: Astro plus a CMS built around how you work.

WordPress → Astro →

How it works

A small audit, not a giant consulting project

From the URL to the action list in about a week. No workshops, no 80-slide deck.

Send me your website

I start with the URL and a short description of what the business actually sells and to whom. That is enough to begin.

I define the query set

I agree with you a focused set of customer questions and the AI systems I will check them in. Written down, so you can come back to it.

I analyse the website

I go through content, structure, entities, technical access and structured data. By hand and with tools.

I compare the results

I document what is clear, what is missing and where the website creates ambiguity — with example answers.

You get the action list

Prioritised findings and concrete next steps. I walk you through them by e-mail, or on a call if you prefer.

Pricing

The audit, or the audit with implementation

Net "from" prices. The audit has a fixed scope: one company website and a defined query set. Quote within 24 hours.

Multilingual, multi-brand or content-heavy websites are quoted separately — the query set and the number of templates grow, and I would rather not pretend one price fits everything.

FAQ

Frequently asked questions

What is an AI Search Audit?

It's a focused review of how clearly your website represents your business to AI-powered search and answer systems. I check what those systems currently understand about you, where the website is ambiguous or silent, and what specifically to change — in the content, the structure and the code.

Is this the same as SEO?

No, though there is overlap. Classic SEO asks how a page ranks for keywords. This audit asks whether a machine can answer a customer's question from your website — and puts more attention on entities, direct answers, structured information and how a defined set of AI systems describes the business. Good work here usually helps SEO too, but it does not replace it.

Can you guarantee that ChatGPT will recommend my company?

No. AI systems use changing data, retrieval systems and models outside my control. I can improve the information and technical signals your website provides, and show the difference on a written query set. I cannot guarantee a specific answer or citation — and I would not trust anyone who promises one.

Is it worth it for a small local business?

Often the most, because questions about local services — "who nearby", "what does it cost", "who should I go to" — are exactly what people ask assistants. A small site has less to fix, so the list is shorter and the changes show sooner. If the site is so small the audit has nothing to look at, I will say so before quoting.

Do I need a new website?

Not necessarily. Sometimes the existing site has a few structural problems that can be fixed in a week. Sometimes the CMS or the frontend is the limiting factor — content rendered by script, crawlers blocked, no way to add structured data. The audit is there to tell you which situation you have before you spend money on either.

Do you need access to my website?

Usually I start from the public website and that is enough for the audit. Deeper technical checks — crawler logs, Search Console, CMS settings — need access, but I only ask for it once I know it will add something.

Which AI systems do you check?

The set depends on the project. Usually ChatGPT, Google AI (Overviews and AI Mode), Perplexity and Gemini, because those are what the customers of the businesses I work for use. I define a focused set around the tools and questions relevant to your audience rather than pretending every model behaves the same way. The list is written in the report.

Can you implement the recommendations?

Yes. I can handle the technical and content changes on the existing site or, when the foundation is the problem, help rebuild it on a more suitable stack. The "audit + implementation" bundle exists for that. No obligation: the report is written so it can be implemented without me.

What does the audit not cover?

I audit the website, not the whole internet. Directory listings, reviews, industry profiles and links from other sites are checked only as far as the AI answers draw on them — I do not build them and do not sell "authority building". There is no ongoing monitoring either: the audit is a snapshot on a date, with the query set written down, so it can be repeated in six months and compared. And there is no 0–100 score — a number that looks serious and says nothing about what to change.

How long does it take?

The audit of a typical company website takes about a week from agreeing the query set. Implementation depends on the list — most often another week or two. The date comes with the quote.

Before you optimise for AI, find out what AI can already understand

Send me your website and the kind of customers you want to reach. I'll tell you what I'd check first and what an audit would cover.

Check your website

The address and two sentences about the business are enough to start. Quote and date within 24 hours.

What do you want checked?
Details and contact

No generic AI strategy deck. Just a practical review of your website. I respond within 2 business hours.