LLM Visibility Monitoring

Record AI visibility over time.

We keep a dated record of selected customer questions, AI answers and sources shown. You can see what changed and whether the facts are right, without treating a small sample as a score for all of AI search.

See what this covers

Where customers can get the wrong picture.

These are possible problems to investigate, not findings from a client audit. Each can lead you to the wrong conclusion about how AI answers describe your business.

01

The question changes between checks.

A small change in wording, tool, mode or location can change the answer. Without those details, two checks may not be comparable.

02

Only brand mentions are counted.

An answer can name you but get your services or service area wrong. Another may not name you, yet still link to your page. Counting mentions alone hides both.

03

A single changed answer is called a trend.

A few checks cannot show what every customer sees. Account settings, such as saved memory or location context, can also affect the answer.

04

A page edit is credited for an answer change without evidence.

A page edit and an answer change can happen close together for unrelated reasons. The dates show timing, not cause.

Make the next decision easier.

We check agreed questions with their conditions recorded, so each observation helps you decide what to investigate next.

01

Keep the conditions with every check.

We record the question, tool, mode, language, available location context and date. Mode is the search experience used, such as Google AI Overviews, AI Mode or ChatGPT Search. Conditions we cannot hold constant remain recorded as limits.

Customer questionRecorded conditionsSourcesNext check recorded
02

Compare factual accuracy and sources.

We compare services and service-area facts, not only whether the answer names or links to your business. A wrong fact and a missing link lead to different investigations.

Approved factsPage taskReviewer
03

Flag changes that need another check.

When an answer changes without a clear reason, we recommend a repeat check before calling it a trend. A factual concern goes to your named approver; the observation alone does not establish cause.

StartDated record
ReviewRepeat check
NextLimits

Follow one customer question through the review.

We follow selected questions through agreed repeat checks and review the dated record for factual or source changes. The steps below show the method; they are not a captured client result.

Office-cleaning planning notes, plain calendar sheets and a blurred laptop prepared for repeat reviews.
  1. 01Keep the customer decision in view.

    Your team owns the business decision; the selected question reflects what your customer needs to know.

  2. 02Agree a comparable watchlist.

    We agree question wording, tool, mode, languages, location context, number of observations and review frequency. An existing AI Visibility Audit can provide a starting list if you have one.

  3. 03Record the dated answer and sources.

    We record missing outputs, visible links and any account or product differences, so the conditions of each check stay clear.

  4. 04Compare facts as well as mentions.

    We compare service-area and service facts with approved information. A wrong mention and a correct source without a mention remain different observations.

  5. 05Review uncertainty and next actions.

    We recommend a repeat check for unexplained changes and pass factual problems to your named approver for source investigation. Edit and answer dates are recorded without assuming causation.

Illustrative work scenario; not a client meeting or captured AI answer.

Agree what this review covers.

Open each area to see a possible scope. We agree available evidence, outputs, responsibility and limits with you before work starts.

01

Possible scope 1

Question watchlist and check conditions

An agreed list of customer questions and recorded conditions.

See what this covers

We agree wording, tools, modes, languages and location context for each question, the number and frequency of checks, and conditions that cannot be held constant.

  • Agree questions and tools in scope
  • Name who approves watchlist changes
02

Possible scope 2

Dated observations and source links

A dated log of answers and sources shown.

See what this covers

Each entry keeps its date, answer wording, visible links and missing answers or sources with the recorded conditions.

  • Agree the evidence format
  • Record unavailable outputs and sources
03

Possible scope 3

Accuracy and mention comparison

Factual and mention comparisons against approved business information.

See what this covers

We compare the answer with approved services, service area and other relevant facts. A correct source without a mention and a mention with a wrong fact remain different findings.

  • Provide approved facts
  • Name who confirms disputed facts
04

Possible scope 4

Repeat-check notes and next actions

Notes on changes, uncertainty and the next investigation.

See what this covers

We note whether an unexplained change needs another check. Page-edit and answer-change dates remain separate; timing alone does not establish cause.

  • Name who investigates a factual concern
  • Keep page changes a separate scope decision

Keep the question, evidence and next step connected.

This is the proposed order of work. Your agreed scope names who handles each step.

  1. 01

    Agree the watchlist and sampling limits.

    We agree the decisions the record should support, questions, conditions, number of checks and review frequency, and what the sample cannot show.

  2. 02

    Repeat and record the checks.

    We run agreed checks and add dated answers, sources and missing outputs to the record with their conditions.

  3. 03

    Review changes without assigning unsupported credit.

    At agreed reviews, we summarise changes, customer impact and uncertainty. Factual concerns go to your approver; page edits remain a separate decision.

See how the question becomes a practical task.

Open the three steps below to see one customer question, the evidence needed and a possible next task. This is an illustrative walkthrough, not a captured answer or client result.

Illustrative working example · Dated observations and review

Prompt watchlist and observation log

We repeat agreed customer questions and compare a dated record of facts, sources and conditions. For example, a location question may reveal that a source describes an outdated service area.

Customer question

What should I check before choosing a supplier?

Repeated questions can reveal factual or source changes, but a small sample cannot establish what every customer sees.

Open each step to see the proposed checks, evidence and possible next task.

01Agree what should stay comparable

Watchlist setup

We agree exact question wording, tool, mode, languages, location context, observation count and review frequency. We note device and account conditions where relevant and available, and differences we cannot hold constant.

02Compare facts as well as mentions

Change review

We record mentions, visible links, missing outputs and whether business facts are correct. A correct source without a mention and an incorrect mention are different observations; the record keeps both the conditions and limits clear.

03Assign the next action honestly

Review note

We assign factual problems for source investigation and repeat unexplained changes before interpreting them. Edit and answer-change dates remain separate: timing alone does not establish causation. Any implementation, or checks beyond the agreed number, are agreed separately.

Platform guidance: Google’s guidance for AI Search features.

What this shows: An agreed scope can provide dated checks and a review summary: what changed, why it matters, what remains uncertain and who should act next.Illustrative process only. No client result, captured AI response or placement is represented.

Working with Aaron.

Bring the customer question, relevant pages and the person who can confirm the facts. Before starting, agree the scope with Aaron, including who approves changes and who implements them.

Aaron reviewing papers during a discussion.
01

Bring the business context.

Tell us what your customer needs to decide and share the page or answer that concerns you.

02

Identify who can verify the facts.

Name the subject expert or business owner who can confirm service area, specifications and scope.

03

Agree implementation and follow-up.

Confirm whether the engagement supplies recommendations, edited pages or implementation, and how follow-up is checked.

Check whether the scope matches your starting point.

Select the statements that describe your situation. This helps frame a discussion; it is not an automated qualification or visibility score.

What can you bring to the work?

Questions about visibility monitoring.

Short answers on what this work can show, what it cannot, and who does what.

Can one changed answer establish a trend?

No. We record the conditions and recommend a repeat check before interpreting an unexplained change. Account, product, mode and date differences can affect comparability.

Does a later answer change prove a page edit worked?

No. An edit and an answer change can happen close together for unrelated reasons. We record both dates and the observed difference, treating the timing as a reason to investigate rather than proof of cause.

How often should checks run?

There is no universal cadence. Agree the question set, number of observations and review frequency around the decisions the record should support.

Is implementation included?

Not automatically. Monitoring records and reviews answers; it does not itself change pages. The proposal names deliverables, factual approver and implementation owner. Any Content Optimisation or page changes are agreed separately.

What access is needed?

We start with relevant public pages and available observations. Agreed Search Console or analytics access can add context. Google’s Generative AI performance report, when available, shows combined AI Overviews and AI Mode impressions by page, country, device and date. Its current documentation lists impressions only, not the questions people asked or a separate click count. We record unavailable checks.

What results are guaranteed?

None of these checks can guarantee inclusion, citations, traffic or enquiries. Agree the deliverables and how they will be reviewed, while keeping business outcomes separate from source observations.

What should remain fixed between monitoring checks?

We keep the agreed question wording fixed and record tool, mode, language and location context. We also log the date and any account or product differences that cannot be held constant, so comparisons have a clear basis.

Can monitoring cover more than one tool or market?

Yes, if the scope identifies separate question sets and records for each tool and market. We compare like with like before combining findings; a larger collection still does not represent every customer’s experience.

What counts as a useful reason to investigate a change?

A wrong service-area statement, conflicting service fact or changed source can justify investigation. We record customer impact and supporting evidence. A new mention alone does not establish business improvement.

Should visibility observations be reported with enquiries and sales?

They can appear in the same review as separate measures. Sampled answers stay distinct from impressions, website visits, enquiries and sales. We do not attribute a business result to a sampled answer without supporting evidence.

A practical starting point

Choose the questions worth watching.

Share the customer question and the source page or observation that needs attention. Confirm the proposed scope before work begins.

What to expect from a call