How our AEO process works: the Four-Signal Method

Most businesses have no idea what AI assistants say about them. Each month we test it, fix the four signals that decide whether you are named, and show you the raw answers.

The Four-Signal Method

Four things decide whether AI names you.

AI assistants build answers from sources they can find and verify. These four signals decide whether yours count. We audit all four, fix the weak ones and track them monthly.

  1. 1Proof
  2. 2Presence
  3. 3Answers
  4. 4Tracking
  1. Signal 1

    Proof

    Proof: Reviews, credentials and mentions on trusted sites. AI leans on what it can verify.

    • Ask happy customers for honest reviews, with a simple and compliant process.
    • Show your qualifications, licences and registrations clearly on your site.
    • Earn mentions on trusted local and industry sites where your name belongs.
  2. Signal 2

    Presence

    Presence: One consistent version of your business on every source AI reads, so it sees one business, not several.

    • Fix mismatched name, address, phone, hours and services across your site, Google Business Profile and directories.
    • Add structured data (hidden labels that help machines read your facts) that matches what visitors can see.
  3. Signal 3

    Answers

    Answers: Pages that answer the exact questions customers ask, in the first sentence.

    • Build your question set: what customers really ask AI, in their own words.
    • Write plain-English answer pages. Each one opens with a one-sentence answer. You approve every page.
    • Share prices or ranges where you can. AI quotes what it can find.
  4. Signal 4

    Tracking

    Tracking: The same question set, re-asked monthly, with dated raw answers.

    • Re-ask ChatGPT, Gemini and Perplexity the same question set every month.
    • Save the raw answers with the date, so you see what AI said, not only a score.
    • Send a plain-English report: what changed, what we did, what is next.
Know the terms

SEO, AEO and GEO: what each one does.

The terms overlap and are often used loosely. This is how we use them.

SEO

Search engine optimization earns a place in Google’s list of results. It is the foundation, because assistants read much of the same web, but a ranking is not the same as being named in an answer. SEO services

AEO

Answer engine optimization makes your business the clear, verifiable source an assistant can quote when it answers a question directly. This is our core service. AEO services

GEO

Generative engine optimization looks at how generative tools describe and compare brands in their own words, including accuracy and tone. GEO services

The monthly process.

The first month adds a full baseline test, so there is a fixed starting point to measure from.

  1. Days 1–3

    Test

    We put your question set to ChatGPT, Gemini and Perplexity in fresh sessions and save every answer, dated.

  2. Days 4–7

    Diagnose

    We record who gets named, which sources the answers cite where shown, and which signal is weakest.

  3. Weeks 2–3

    Fix

    We fix listings, gather proof and write answer pages. You approve before anything goes live.

  4. Week 4

    Report

    You get the raw answers, what changed, what we did and what comes next. A video call is optional.

Why the same questions every month? AI answers vary from one session to the next. Re-asking a fixed set, in fresh sessions, on a regular schedule is the most honest way to tell a real change from noise.

Inside the monthly test.

How we keep the test honest, and how a test result turns into a list of fixes.

Fresh sessions, every time

Each question is asked in a new session, so earlier conversations cannot shape the answer. Assistants still vary from one session to the next, which is why we treat one answer as a data point, not a verdict.

Recording, not just scoring

We save the full text of every answer with its date. A score is useful for spotting trends, but the wording matters too: an assistant can name you and still describe you inaccurately, which is the territory of generative engine optimization.

Reading the sources

Where an assistant shows its sources, we log them. If the same sites keep appearing, they become priorities: a listing to correct, a directory to join or an organisation worth contacting. For Google’s own AI features, see AI SEO.

Choosing what to fix first

We rank fixes by how directly they relate to the questions where you are missing, and by the effort involved. A wrong address on a frequently cited directory comes before a nice-to-have article. Listing fixes often start with your Google Business Profile.

What you receive each month.

One report, in plain English, with nothing hidden behind a dashboard login.

  • The raw answers from ChatGPT, Gemini and Perplexity for every question in your set, with dates and any sources shown.
  • A question-by-question table: named, not named, or named but described inaccurately, compared with last month.
  • The businesses named in your place, so you can see who you compete with in AI answers.
  • A log of every change we made: listings, pages, structured data and outreach.
  • Drafts waiting for your approval, and the plan for next month.
  • An optional short call to walk through it.

Example: the first three months.

A hypothetical timeline for an invented business, to show how the work unfolds. Real timing varies by business and by assistant.

Example

Month one: baseline and the obvious fixes

We agree the question set, run the baseline and find that the business is named for two questions about its main service but none about its second service. Two directory listings show an old phone number. We correct the listings, draft an answer page for the second service and send both for approval.

Example

Months two and three: answers and proof

The new page goes live. We add practitioner credentials as text, help the owner ask recent customers for reviews and contact a local association that assistants cited. Each re-test is compared with the baseline, question by question, and the report says plainly what moved, what did not and what we will try next.

See how this applies in a sector: aesthetic clinics, law firms and home improvement contractors.

Common mistakes when measuring AI visibility.

  • Changing the questions every month. If the test changes, you cannot tell whether a different answer reflects a real change.
  • Testing while signed in. Personal history and earlier chats can shape answers, so results may not reflect what a new customer sees.
  • Counting mentions without reading them. Being named with the wrong address or an outdated description can do more harm than not being named.
  • Treating AI visibility as separate from search. Assistants read much of the same web as search engines. A site that cannot be crawled holds both back; technical SEO fixes that.
  • Expecting overnight change. Assistants and search indexes refresh at different speeds. Judge the trend over months, not one re-test.

Getting started takes about two minutes.

You see the test results before you spend anything.

  1. Tell us about your business

    Name, website and city on the free AI test form.

  2. We run the test

    We ask three AI assistants the questions your customers ask and save the answers.

  3. You get a scorecard

    Within 7 days: who was named, whether you were, and your top three fixes.

  4. You decide

    Make the fixes yourself, or ask us to run the monthly work. There is no follow-up pressure.

What we will not do.

  • Write, buy or incentivise fake reviews.
  • Promise a ranking, a mention or a number of new customers.
  • Publish anything you have not approved.
  • Give legal, medical, financial or other regulated advice.

Common questions

Which AI assistants do you test?

ChatGPT, Gemini and Perplexity. If your customers rely on another assistant, tell us and we can discuss adding it.

What is a question set?

The real questions your customers ask, in their words, such as “best aesthetic clinic in Leeds” or “who handles a partner dispute”. We build it from your services and area at the start and re-ask the same set every month.

Do you write reviews for me?

No. Fake or incentivised reviews break platform rules and can harm you. We help you ask real customers for honest reviews in a way that follows the platforms' rules.

What if AI already names me?

Then the work is about protecting that position and finding the questions where you are still missing or described inaccurately.

Do I need to drop my current SEO or marketing agency?

No. We can work alongside them and share what the tests show. Much of what helps AI visibility also helps search.

What do you need from me?

About 30 minutes a month to approve work, access to your website or someone who can publish changes, and accurate details about your services, credentials and service area.

How many questions are in a question set?

It depends on how many services and areas you cover. We keep the set small enough to test carefully every month and broad enough to reflect what your customers really ask.

Can I see the raw answers myself?

Yes. Every report includes the full answers from each assistant, with dates, so you can read exactly what was said rather than relying on our summary.

Find out if AI names you.

Get your free AI test. We send the results in 7 days.