Visibility baseline
Best for
Organisations that do not yet know how assistants describe them.
Typical output
Baseline results, citation map and an accuracy log.
AI search visibility
Being findable in Google no longer means being present in the answer.
AI assistants increasingly answer the question instead of listing the links. We test how they currently describe your organisation, which sources they rely on, and what it would take to be cited rather than omitted.
Working method
One good answer proves nothing. The same prompts, run again, do.
- 01 Baseline Run the stored prompt set across the assistants your buyers actually use, and record every answer and citation.
- 02 Diagnose Separate a visibility problem from an accuracy problem. Being absent and being described wrongly need different work.
- 03 Strengthen Correct the third-party sources assistants lean on, and make your own pages easier to extract and quote.
- 04 Re-test Run the identical set again and compare. Movement across repeated runs is evidence; a single run is anecdote.
Engagement options
Start at the smallest level that can answer the decision
The scope is agreed around the workflow, people, systems and evidence required. You do not need to commit to a large programme before a focused diagnostic or pilot has earned it.
Baseline plus plan
Best for
Teams ready to act on the gap once it is quantified.
Typical output
The above, plus a prioritised source-authority plan.
Implementation
Best for
Organisations that want the content and technical work carried out, not just specified.
Typical output
Implemented changes, with before and after evidence.
Ongoing monitoring
Best for
Categories where assistant answers are shifting month to month.
Typical output
Scheduled re-tests and a change log against the baseline.
Evidence standard: Assistant answers vary between runs, so a single screenshot proves very little. Every baseline is built from a stored prompt set run repeatedly, and every claim of improvement is evidenced by re-running that same set rather than by a one-off comparison.
Founder-led and grounded in the room
The questions people actually ask shape the session.
Jon works directly with the team, using its roles, documents and live challenges to make the material useful beyond the workshop.
The aim is not tool excitement. It is a shared method people can repeat, review and improve once they are back at work.
Built around real work
A useful programme, not a generic tool tour
The exact examples change by team, but the programme is grounded in the following practical areas.
Assistant baseline testing
Structured, repeatable prompts across the assistants your buyers actually use.
- Buyer-intent questions in your category
- Comparison and alternative-to questions
- Whether you are named, described correctly or omitted
Citation and source analysis
Which pages and third-party sources assistants rely on to answer.
- Your own pages that are and are not being drawn on
- Third-party sources carrying weight in your category
- Where a competitor owns the cited source and you do not
Technical and structural foundations
Whether your site can be crawled, parsed and attributed correctly.
- Crawl, indexation and rendering behaviour
- Entity, schema and organisational markup
- Content structure that supports direct extraction and quotation
Source-authority plan
A prioritised route to being cited rather than merely present.
- Pages to create, strengthen or correct
- Third-party sources worth earning a presence in
- A re-test method so change can be evidenced over time
How the engagement works
- 1
Define the questions that matter
Agree the buyer-intent questions where being cited would have commercial consequence.
- 2
Baseline and record
Run those questions across the assistants, capture the answers and the cited sources, and store the results so they can be compared later.
- 3
Diagnose the gap
Separate problems of visibility, accuracy and attribution. Each has a different fix.
- 4
Implement and re-test
Make the content, technical and source changes, then re-run the same prompts to evidence movement.
What you take away
- AI visibility baseline across the assistants your buyers use
- Prompt set and results, stored so the test can be repeated consistently
- Citation map showing which sources assistants draw on in your category
- Accuracy log of anything assistants state about you that is wrong
- Prioritised source-authority plan with named pages and third-party targets
- Re-test schedule and the measures that would show genuine improvement
Read before you decide
Technical SEO audit checklist
The crawl, indexation and structural foundations that AI search visibility depends on.
Read the practical guideNeed ongoing learning rather than a private team engagement? Explore Jon’s AI Essentials learning pathway in a new tab.
Questions teams ask
What is AI search visibility?
It is whether AI assistants such as ChatGPT, Claude and Gemini can find your organisation, describe it accurately, and cite it when answering questions in your category. It overlaps with SEO but is not the same discipline: the unit of success is being cited in an answer rather than ranking in a list of links.
How is this different from technical SEO?
Technical SEO makes a site crawlable, fast and correctly structured, and that remains a prerequisite. AI search visibility asks a further question: given that assistants can reach your pages, do they actually draw on them when they answer, and do they describe you correctly when they do. A site can be technically excellent and still be absent from the answer.
Can you actually influence what an AI assistant says?
Not directly, and anyone promising to control an assistant’s output should be treated with suspicion. What can be influenced is the evidence available to it: whether your own pages answer the question clearly and can be extracted, whether the third-party sources it relies on describe you accurately, and whether your organisational entity is unambiguous. The honest framing is influence over the inputs, not control of the output.
How do you measure whether it worked?
By running the same prompt set before and after, storing both sets of results, and comparing whether you are named, how you are described and which sources are cited. Because assistant answers vary between runs, a single before and after comparison is not sufficient; the test needs repeating across runs to distinguish real movement from noise.
Does this matter if we already rank well in Google?
Increasingly, yes. Ranking well protects the click that starts in a search box. It does not protect the recommendation that happens inside an assistant, where a single answer may name two or three providers and omit everyone else. The organisations at most risk are those whose good rankings mask the fact that they are absent from the answer.
What if an assistant says something inaccurate about us?
That is a finding worth acting on quickly, and it is more common than most organisations expect. The fix usually involves correcting or clarifying the underlying sources the assistant is drawing on, and making the accurate version easier to extract from your own pages. We log every inaccuracy found during the baseline.
Related support
Continue with the page that matches the next decision
Technical SEO
The crawl, rendering and structural work that AI visibility depends on.
Explore this serviceAI readiness audit
The internal counterpart: whether AI is worth applying to your own operations.
Explore this serviceDigital PR
Where the gap is third-party sources, earned coverage is often the route to being cited.
Explore this serviceFind out what AI assistants say about you
Tell us about the team, the workflow and what has prompted the conversation. We will recommend a proportionate first step.
Start the conversation