An unbranded AI visibility test asks whether Google independently recommends your business for a particular customer need.
But what if the business does not appear?
That result alone cannot tell you whether Google lacks information about the business, misunderstands its offer or can recognise its suitability when prompted but chooses other businesses in an unbranded search.
A second test can help investigate the difference. Name the business, describe a realistic customer situation and ask AI to assess how well the published evidence matches the customer’s requirements.
For example:
Assess whether ABC Accountants of Birmingham appears suitable for a 25-employee manufacturing company that uses Xero and needs payroll, year-end accounts, corporation tax and monthly management accounts for a fixed monthly fee. Do not assume that the firm is suitable. Evaluate each requirement separately, cite the supporting evidence and identify anything that cannot be confirmed.
This is the Scenario Suitability Test.
It does not determine whether the business is genuinely suitable or whether Google will recommend it naturally. It tests whether AI can construct an accurate, evidence-supported assessment of the business when given a defined customer need.
Discovery and suitability are different questions
The first test should normally be unbranded:
Which Birmingham accountancy firms should a 25-person manufacturer using Xero consider?
That measures discovery. Does the firm enter the recommendation candidate pool without being named?
The Scenario Suitability Test asks a different question:
When the business is named, can AI explain whether its publicly available evidence matches this customer’s needs?
| Test | Question answered |
|---|---|
| Unbranded theme test | Does AI independently discover and recommend the business? |
| Named Scenario Suitability Test | When prompted, does AI find evidence that the business fits the scenario? |
A business may appear suitable when named without being recommended regularly in unbranded searches. That could indicate a discovery or comparative visibility problem rather than a complete lack of relevant information.
It is only a starting point for investigation, however. Stronger competitors, external authority, indexing and the limited test sample may also affect the unbranded results.
Naming the business changes the task
The Scenario Suitability Test introduces an important bias.
Once ABC Accountants is named, AI may search more deliberately for information about that firm and try to produce a helpful assessment. It may assemble evidence that would not have caused the business to appear naturally in an unbranded recommendation.
A positive named assessment therefore does not prove that:
- The business would enter the recommendation candidate pool without prompting
- Google considers it one of the strongest available choices
- The firm would appear for other formulations of the customer need
- The firm is genuinely suitable in practice
The test deliberately removes the discovery stage. Its purpose is diagnosis, not proof of unprompted visibility.
This extends the Interpretation Test
The Interpretation Test asks AI what it knows about a business and compares the answer with the identity the business intended to communicate.
The Scenario Suitability Test adds a defined customer and explicit selection criteria.
- Interpretation Test: Does AI understand our business generally?
- Scenario Suitability Test: Does AI understand how our published evidence relates to this particular customer need?
A general answer might correctly identify ABC Accountants as a Birmingham firm using cloud software. That does not establish whether its website supports the case for choosing it to serve a 25-person manufacturer requiring Xero, payroll and monthly management accounts.
The scenario creates a more commercially useful test.
Assess public evidence—not actual commercial suitability
AI can normally assess only publicly discoverable information. It cannot reliably establish:
- Whether the firm currently has capacity for another client
- Whether the standard package includes every requested service
- What the firm would quote this particular business
- Whether the claimed expertise reflects the current team
- Whether the working relationship would be suitable
- Whether the prospective client would pass the firm’s acceptance procedures
The conclusion should therefore be framed as:
The business appears suitable, partially suitable or unsuitable based on the public evidence found.
It should not be presented as professional advice or a definitive recommendation to appoint the firm.
An accurate negative assessment can also be a successful result. If ABC Accountants does not serve manufacturers or cannot support the required payroll, AI explaining that accurately shows correct interpretation. The aim is not to force every named business into a favourable answer.
Break the scenario into assessable requirements
Before running the test, list the selection criteria contained within the scenario.
For the manufacturing-accountant example:
- Birmingham presence or coverage
- Manufacturing experience
- Suitability for an established 25-employee business
- Year-end accounts
- Corporation tax
- Payroll capacity
- Monthly management accounts
- Xero expertise
- Fixed monthly fees
Then use a consistent classification for each requirement:
| Classification | Meaning |
|---|---|
| Supported | Direct and reliable evidence confirms the requirement |
| Partially supported | Relevant evidence exists, but an important detail is missing |
| Unconfirmed | No reliable public evidence was found |
| Contradicted | Available evidence suggests the firm does not meet the requirement |
| Incorrectly interpreted | AI made a claim that its source does not support |
This prevents the test from becoming a vague judgement about whether the firm “looks suitable.”
Start with a neutral prompt
Avoid leading wording such as:
Explain why ABC Accountants is the best choice for this manufacturer.
That encourages AI to construct a favourable case.
The neutral prompt at the start of this article does four useful things:
- It says not to assume suitability.
- It requests a separate assessment of every requirement.
- It asks for supporting evidence.
- It asks AI to state what cannot be confirmed.
It may be better to request the requirement-by-requirement assessment before asking for an overall conclusion. This reduces the pressure to decide “yes” and then justify that answer retrospectively.
Ask about evidence, not hidden reasoning
After the initial assessment, useful questions include:
- What evidence demonstrates the firm’s manufacturing experience?
- What proves that it has genuine Xero expertise?
- Does the website confirm the size of payroll it can support?
- Are monthly management accounts explicitly offered?
- Does “fixed fee” mean a continuing fixed monthly package?
- Which pages provide the strongest evidence?
- Which requirements remain unconfirmed?
- Does the evidence come from the firm or independent sources?
- Are there case studies or testimonials from comparable customers?
These questions concern information that can be checked.
“Why didn’t you recommend ABC Accountants earlier?” is less reliable. AI may generate a plausible retrospective explanation rather than reveal the retrieval, weighting or recommendation process that produced the earlier answer.
Google’s hidden query fan-out and internal decision-making are not exposed through a follow-up question. The safest approach is to examine claims, citations and missing evidence—not treat AI’s explanation as access to its ranking logic.
Use fresh sessions for independent checks
Follow-up questions within one conversation are helpful for exploring an answer, but they are not independent tests. Later responses may inherit assumptions from the initial assessment.
A stronger method combines:
- One conversational investigation: Ask the main suitability question and explore the answer with follow-ups.
- Fresh-session repetitions: Run the main prompt several times in separate conversations.
- Independent evidence checks: Ask important requirement questions separately so they are not anchored by the first verdict.
Keep the platform, location conditions, wording and test period as consistent as possible. Record the date, answer, cited pages and material claims.
Several runs are more informative than one. A single confident assessment may not be stable.
Build a requirement-by-evidence table
The results can be recorded like this:
| Requirement | AI assessment | Evidence found | Manual verification |
|---|---|---|---|
| Birmingham | Supported | Birmingham office page | Accurate |
| Manufacturing | Partially supported | Manufacturing listed as a sector | General claim only |
| 25-person payroll | Unconfirmed | Payroll service page | Capacity not stated |
| Monthly management accounts | Supported | Management reporting page | Accurate |
| Xero expertise | Partially supported | Xero logo and service page | Certification not confirmed |
| Fixed monthly fees | Supported | Pricing page | Accurate |
| Comparable results | Unconfirmed | No relevant case study found | Evidence gap |
This exposes the difference between mentioning a service and demonstrating expertise.
A Xero logo may establish an association with the software. It does not prove how many certified advisers the firm employs, how many migrations it has completed or whether it can configure useful reporting for a manufacturer.
Compare discovery with named suitability
Combining the unbranded and named tests creates several useful investigative questions:
| Unbranded result | Named assessment | Initial question to investigate |
|---|---|---|
| Appears regularly | Requirements well supported | Are discovery and public evidence both comparatively strong? |
| Rarely or never appears | Requirements well supported | Why is apparently strong evidence not producing regular discovery? |
| Appears regularly | Important requirements unconfirmed | What other signals may be supporting the recommendation? |
| Rarely or never appears | Important requirements unconfirmed | Is the public evidence too weak, general or difficult to find? |
| Appears but is described inaccurately | Conflicting assessment | Is AI using mistaken or inconsistent information? |
These are questions, not diagnoses. A named assessment cannot reveal Google’s ranking formula.
Look for claim, proof, application and outcome
The test can reveal whether a website merely makes claims or provides a credible case for believing them.
| Layer | Xero example |
|---|---|
| Claim | We are specialist Xero accountants |
| Proof | Current partner status, certified advisers and completed migrations |
| Application | Xero reporting, payroll and inventory integrations for manufacturers |
| Outcome | Faster reporting and clearer stock, margin and cash-flow information |
A claim tells customers what the firm says it can do. Proof supports that claim. Application connects the expertise with the customer’s situation. Outcomes show what changed for comparable clients.
Every statement must be accurate and supportable. The purpose is not to manufacture credentials for AI visibility, but to make genuine expertise easier to verify.
Verify every material claim and source
The AI assessment is not the final evidence. It is part of the test.
For every significant claim, check:
- Whether the cited page actually supports it
- Whether the information is current
- Whether AI has confused similarly named businesses
- Whether a certification or directory listing remains valid
- Whether “management accounts” explicitly means monthly reporting
- Whether an industry list demonstrates expertise or merely claims coverage
- Whether the firm’s website and external sources agree
External sources should also be classified. A current Xero directory profile, detailed customer case study or independently verified award may provide valuable corroboration. An old directory listing or generic review may be much weaker.
Record the source type, date where relevant, who controls it and whether it directly supports the requirement. This follows the same principle as checking whether AI citations genuinely support the claims being made.
An incorrect but favourable answer is not a successful result. It shows that AI’s interpretation is unreliable or that the available evidence is ambiguous.
Classify the evidence problem
The test may reveal several different weaknesses:
| Evidence problem | Meaning |
|---|---|
| Missing | The business provides the service but does not state it publicly |
| Weak | A claim is made without convincing proof |
| Scattered | Relevant evidence exists across disconnected pages |
| Ambiguous | Important details such as frequency, scale or package contents are unclear |
| Inconsistent | Different pages or external sources conflict |
| Externally dependent | AI relies mainly on third-party sources to understand the business |
The correct response is not automatically to write a longer page. It may involve correcting outdated information, connecting existing resources, adding verifiable proof or making an important commercial detail explicit.
An eight-step Scenario Suitability Test
- Choose the customer scenario. Use a commercially important need the business genuinely wants to serve.
- Run the unbranded test first. Record whether the business is independently recommended.
- List the requirements. Break the scenario into assessable selection criteria.
- Run the neutral named assessment. Repeat it in fresh sessions and request evidence and unconfirmed points.
- Probe important requirements. Use separate questions where the initial evidence is weak or ambiguous.
- Record and verify everything. Check every material claim, citation and external source.
- Compare discovery with suitability. Identify gaps between unbranded visibility and named understanding.
- Improve genuine evidence and repeat. Allow meaningful changes to be indexed before running the same controlled test again.
Can AI make an accurate case for choosing you?
Brand visibility is not only about whether a business name appears.
A commercially useful recommendation requires AI to understand who the business helps, what it can do and why its claims should be believed.
The Scenario Suitability Test asks:
When AI is given a realistic customer need, can it build an accurate, evidence-supported assessment of our business from public information?
If it can, but the business rarely appears in unbranded searches, the next investigation concerns discovery and comparative visibility.
If it cannot, the website may not provide enough clear evidence for AI—or a prospective customer—to understand the fit.
If it accurately concludes that the business is unsuitable, that is useful too. Correct interpretation is more valuable than an unsupported recommendation.
One result may reveal a visibility gap. Another may reveal an evidence gap. Both give the business something specific to investigate.