How to Check Whether Gemini Recommends Your Business: A Worked UK Example
How to Check Whether Gemini Recommends Your Business: A Worked UK Example
If a prospective customer asks Gemini for a supplier like yours, does your business appear? The honest way to answer is not to run one flattering prompt and take a screenshot. It is to define the buying questions, record the conditions and keep recommendations separate from citations, visits and enquiries.
We ran a small baseline for Byte Digital on 2 November 2026. It did not produce the result we wanted, which is exactly why it is useful.
The three questions we tested
We used these exact prompts:
- Which UK AI SEO agencies would you shortlist?
- Can you recommend a GEO agency in London?
- I need help with AI SEO for a UK service business. Where should I start?
Each prompt was run once in its own signed-in UK temporary Gemini chat, using Auto mode. We did not name Byte, provide our URL, request particular sources or regenerate the answer. The interface did not display the underlying model version.
This is a baseline, not market research. Gemini responses can vary by account, date, product mode and other conditions we cannot see. A single run tells us what appeared in that run. It does not tell us what every UK customer sees.
What appeared in the answers
The first answer shortlisted UK agencies and distinguished GEO specialists from agencies using AI within SEO. Byte was not named.
The London answer named five agencies. Byte was not among them. Source cards were attached to the response.
The third answer was different. It offered a practical starting sequence for a UK service business, covering a visibility baseline, business details, schema, service content and external evidence. It cited several guides, but no Byte page.
The counts therefore need to stay separate:
Combining those into “0 out of 3 rankings” would make the record sound simpler while making it less accurate. Two prompts asked for suppliers. One asked for advice. A recommendation and a source citation are different outcomes.
What we could learn from the visible sources
In the London answer, Gemini linked to Iconic Digital’s GEO service page . The page gave the system a direct account of the agency’s service and positioning.
The same answer used Blue Array’s UK GEO agency directory in connection with POLARIS. That directory is published by an agency and includes its publisher, so it is not independent verification of commercial results. It still gave the answer a page that explicitly connected named suppliers with the category.
For the practical-help question, Gemini cited tektiv’s guide for small businesses . The guide includes an actionable starting section aimed at that audience.
These observations show what the visible source cards contributed. They do not reveal why Gemini selected those companies or pages. Google’s own Gemini Apps source guidance says that sources and related links may appear within or below a response, and that not every response provides them.
Gemini is not the same as Google AI Mode
This test covered the consumer Gemini web app. It did not cover Google AI Overviews, AI Mode or an API response.
That distinction matters because Google describes AI Overviews and AI Mode as features of Google Search, grounded in its Search index and ranking systems. Its guidance for generative AI features in Search should not be used to relabel a Gemini app observation as a Google Search ranking.
If you test several surfaces, record them in separate columns. Otherwise a change in one product can be mistaken for progress everywhere.
A test record you can copy
Use one record for every answer:
| Field | Record | | — | — | | Question | | | Date and time | | | Product and mode | | | Signed in or signed out | | | Country or location shown | | | Named companies | | | Was our business recommended? | Yes / No | | Cited or linked URLs | | | Was our page cited? | Yes / No | | Important factual errors | | | Screenshot or retained evidence | | | Next action | |
A miss belongs in the record. Do not delete it because the answer is inconvenient, and do not keep regenerating until the brand appears. That turns a baseline into a search for a screenshot.
How to repeat the test
Repeat the original wording in fresh chats on separate days. Keep the sign-in state, market and mode as comparable as the interface allows. Record the answer once, including absences and visible source URLs.
Then compare:
Do not call a citation an enquiry. Do not call an appearance a recommendation unless the answer actually recommends the company. And do not treat your own verification visits as customer demand.
Our AI search measurement guide explains that wider measurement ladder. If you want to examine how Byte approaches the work commercially, see our AI SEO and generative engine optimisation service.
What we would do next
The baseline gives us a short list of work to inspect, not a licence to copy the cited pages. We would review whether Byte’s service page answers the buying question clearly, whether our relevant evidence is easy to follow and whether useful third-party records accurately describe the business.
We would then document any page changes before repeating the same prompts. Without that change log, a later appearance might be interesting but impossible to interpret.
FAQs
How many Gemini tests are enough?
One run is enough to create an honest starting observation, but not enough to estimate how often a business appears. Repeat the same questions on separate days and keep the conditions as consistent as possible. Report the number of runs rather than implying a larger sample.
Should I test prompts that include my company name?
Branded prompts are useful for checking factual accuracy. They are not a substitute for unbranded buying questions, where the customer has not already chosen you. Keep branded accuracy checks and unbranded supplier-discovery tests separate.
Does a source card mean Gemini recommends the source?
No. A source card indicates a related or supporting link in that response. The page may support a factual point or practical step without its publisher being recommended as a supplier.
Can I compare Gemini results with AI Overviews?
You can compare observations, but label the surfaces separately. The consumer Gemini app, Google AI Mode and AI Overviews are different products and experiences. A result in one should not be reported as a ranking in another.
What should I do when my business is absent?
Record the miss, inspect the pages and evidence used by the visible answer, and audit the clarity of your own relevant page. Make supportable improvements for buyers, then repeat the original test later without changing the question to favour your brand.
Oops! We could not locate your form.
Want a documented starting point rather than a collection of screenshots? Request Byte Digital’s free initial AI visibility review.
Next step: Request a free initial AI visibility review
Gareth Allen is the founder and Managing Director of Byte Digital, with more than 26 years’ experience across digital marketing, design and web development. He helps businesses use paid media, websites and conversion-focused strategy to generate more qualified leads and sales. His work spans Google Ads, social advertising, campaign planning, tracking and website optimisation. With a background in design and front-end development, Gareth combines commercial marketing expertise with a practical understanding of how websites perform — from user experience and page speed to conversion tracking and landing-page improvements.