UK Retail AI Visibility Benchmark: Citation Share and Competitor Position

Benchmark status

This article sets out a transparent framework for measuring UK retail AI visibility, including citation share, recommendation position, source coverage, and competitor visibility in AI-generated answers. It does not publish brand rankings or performance claims, because a complete, reproducible dataset is not included here.

That distinction matters. A credible AI visibility benchmark should show not only the outcome, but also the prompts tested, platforms used, collection dates, repeat runs, scoring rules, and quality checks behind it.

AI-generated answers are becoming another route through which people discover information, compare options, and build shortlists. Ofcom describes generative AI search tools as “answer engines” because they generate responses rather than simply directing people to webpages. Its UK research also reports growing use of AI services and the wider appearance of AI summaries in search experiences. (ofcom.org.uk)

For UK brand, ecommerce, and insight teams, the commercial question is simple: when shoppers ask an AI system what to buy, which brands are mentioned, which retailers or publishers are cited, and where does each competitor appear?

Why this matters for retail teams

Traditional search reporting usually focuses on rankings, clicks, and impressions. AI-generated answers create a different visibility challenge. A brand may be relevant to a shopper’s question without appearing in the answer at all. Another may be recommended prominently, while a third gains visibility only because its product page, review, or retailer listing is cited as evidence.

This creates four practical questions for retail decision-makers:

• Discoverability: Is the brand appearing in non-branded shopping questions?
• Recommendation share: How often is it included in the shortlist or named as a suitable option?
• Evidence strength: Which websites and pages are being used to support the answer?
• Competitive position: Is the brand appearing early, late, or not at all compared with its competitors?

The measures in this benchmark are designed to separate those questions rather than hide them inside a single proprietary score. That matters because existing AI search visibility tools use different prompt databases, platform coverage, and scoring approaches. For example, Semrush describes separate measures for mentions, citations, cited pages, share of voice, and average position. (semrush.com)

This benchmark therefore treats an AI answer as an observed research event, not as a conventional search-ranking result. Each answer is captured, preserved, scored, and linked to the conditions under which it was generated.

How the benchmark is built

The following design represents the minimum standard for a publishable UK retail benchmark.

Scope

The benchmark covers consumer-facing UK retail and FMCG categories, including grocery, beauty, fashion, home, electronics, and health and wellness products where shopping recommendations are appropriate. Results should be reported by category and platform, not only as an overall total.

Sample period

Every published result should include an exact collection window, such as 1 April to 30 June 2027. Without a collection date, a result is not a reliable benchmark, because AI answers, source indexes, product availability, and platform behaviour can change.

Buyer-question categories

Prompts should reflect realistic UK shopping journeys rather than brand-led keyword lists. Recommended categories include:

1. Category discovery: “What are the best moisturisers for sensitive skin in the UK?”
2. Need-state comparison: “Which cordless vacuum is best for a small flat?”
3. Budget-led shopping: “What are good running shoes under £100 in the UK?”
4. Attribute-led selection: “Which supermarket has the best range of vegetarian ready meals?”
5. Occasion-led shopping: “What should I buy for a UK garden party?”
6. Retailer and availability questions: “Where can I buy a reliable air fryer with fast UK delivery?”
7. Brand comparison: “How does Brand A compare with Brand B for everyday use?”

Each prompt should be assigned one primary category, one product category, a price or need-state label where relevant, and a UK location setting where the platform supports it.

Platforms tested

Results should be recorded separately for each answer environment. A suitable starting set may include ChatGPT with web search, Google AI Overviews or AI Mode where available, Gemini, and Perplexity AI. ChatGPT search responses can include links to web sources, while Perplexity states that its answers include direct source links for verification. (help.openai.com)

Platform names should never be treated as interchangeable. Their retrieval systems, answer layouts, model versions, location handling, and citation behaviour may differ.

Prompt handling and repeatability

Prompts should be written before data collection begins and stored in a locked master list. Researchers should record:

• The exact wording, spelling, and punctuation
• The date and time of each run
• Country, language, and location settings
• Logged-in or logged-out status where relevant
• Platform and model or search mode, if shown
• The complete answer, not only the first paragraph
• All visible citations, links, and source labels
• Any product or retailer availability stated in the answer

Each prompt should be run at least three times on separate dates. The published dataset should show whether a brand appeared consistently, occasionally, or only once. This helps prevent a single unstable answer from being presented as a durable market position.

The scorecard explained

Measure	Plain-English definition	Suggested calculation	Why it matters to retail teams
Citation share	The percentage of all recorded source citations attributed to a brand’s domain or approved source set.	Brand citations ÷ all qualifying citations × 100	Shows whose webpages, product information, or owned content are being used as evidence.
Recommendation position	The brand’s position within the recommended list or answer, counting from the first clearly recommended brand.	Mean position across valid repeated answers	Indicates prominence, not just presence. A brand named first is not equivalent to one mentioned at the end.
Source coverage	The proportion of benchmark prompts for which at least one approved brand-owned page is cited.	Prompts with an owned citation ÷ prompts where the brand is relevant × 100	Shows whether the website is supplying evidence across the buying journey.
Competitor visibility	The proportion of valid answers in which a competitor is mentioned or cited.	Competitor appearances ÷ valid answer runs × 100	Identifies where competitors are present in questions that matter to the category.
Mention rate	The percentage of answer runs in which a brand is named, whether or not it is cited.	Runs mentioning brand ÷ valid runs × 100	Separates brand recognition from source authority.
Consistency	How often the same brand appears across repeat runs of the same prompt.	Runs with the same brand outcome ÷ total repeat runs × 100	Shows whether a result is repeatable or highly variable.

These measures should remain separate. A brand can have a high mention rate but low citation share, or strong source coverage but a weak recommendation position. Combining them without explanation only makes the result harder to interpret.

Results and reading notes

No verified brand results are published in this release

Benchmark results should only be published when another analyst can reproduce the collection and reach the same scoring decision using the stated rules. This article does not include a verified dataset, so it makes no claims about which UK retailer, FMCG brand, or platform currently leads on citation share or recommendation position.

The correct interpretation is therefore methodological: this framework defines what should be measured, how it should be measured, and which evidence must accompany any future ranking.

When verified results are available, the results section should use a table such as the following:

Category	Platform	Prompt runs	Brand	Mention rate	Citation share	Average recommendation position	Source coverage	Consistency
Beauty	ChatGPT search	Verified count	Brand A	Pending	Pending	Pending	Pending	Pending
Grocery	Google AI results	Verified count	Brand B	Pending	Pending	Pending	Pending	Pending
Home	Perplexity AI	Verified count	Brand C	Pending	Pending	Pending	Pending	Pending

Any published chart should show confidence and context, not just a league table. At minimum, it should display the number of prompts, the number of repeat runs, the collection period, and whether the measure is based on mentions, recommendations, or citations.

How to interpret competitor position

Competitor visibility is not the same as market share, sales leadership, or consumer preference. It simply means that a competitor appeared in the sampled AI answers under the stated conditions.

For example, a retailer may appear frequently because it has broad product availability, while a specialist brand may appear less often but in a more prominent recommendation position. A publisher or review site may also receive many citations without being a retailer or product manufacturer.

The benchmark should therefore classify every appearance into at least four groups:

• Brand mention: the brand is named in the answer.
• Product recommendation: a product is actively suggested.
• Owned citation: a brand-controlled page is cited.
• Third-party citation: a retailer, publisher, review site, forum, or marketplace is cited.

This classification makes the output more useful for brand, ecommerce, and category teams. It shows whether the issue is recognition, recommendation, evidence, or third-party coverage.

Comparing AI visibility platforms fairly

Searches for Best AI visibility tools, AI visibility tracker, and AI search visibility tools increasingly return products with different definitions of visibility. Semrush, Peec AI, and Otterly AI may not use the same prompt sources, answer environments, refresh schedules, or scoring logic. A platform score should therefore be treated as that platform’s own measurement system, not as an objective industry currency.

Semrush publicly describes a large prompt database and separate reporting for mentions, citations, and AI visibility. Those measures can be useful for operational monitoring, but they should not be directly compared with another vendor’s score unless the underlying methodology is aligned. (semrush.com)

For procurement or tool evaluation, retail teams should ask:

1. Which platforms and countries are covered?
2. Are prompts fixed, sampled from user behaviour, or generated by the provider?
3. Are answers collected live, cached, or simulated?
4. How are model and platform changes recorded?
5. Are citations counted by URL, domain, or source appearance?
6. Can users export the original answers and collection dates?
7. How are brand variants, sub-brands, and misspellings identified?
8. Are repeat runs and answer variability visible?

The most useful tool is not necessarily the one with the biggest score. It is the one that makes the evidence behind the score inspectable.

Limitations

This benchmark can measure observed visibility in a defined sample. It cannot prove that AI visibility caused a sale, that a cited page was read, or that a recommendation reflects the full UK market.

Key limitations include:

• AI answers can change between runs and across user accounts.
• Platform interfaces, models, retrieval systems, and citation formats can change.
• A finite prompt set cannot represent every UK shopper question.
• Recommendation position may be ambiguous when answers use prose rather than numbered lists.
• Source citations do not necessarily indicate endorsement or product quality.
• Retail availability, price, and delivery information can become outdated quickly.
• Brand names may be confused with products, sub-brands, or similarly named organisations.
• Results collected in one location or language setting may not generalise across the UK.
• Mention and citation metrics do not measure conversion, revenue, or sentiment on their own.

The benchmark should also avoid implying that a website can guarantee inclusion in an AI answer. Search systems use their own retrieval and generation processes, and even cited-answer platforms acknowledge that users should verify important information against reliable sources. (help.openai.com)

Publication standard

A trustworthy UK retail AI visibility benchmark should publish the framework before publishing the league table. The minimum evidence pack includes:

• Exact sample period
• Platform and model details
• UK location and language settings
• Prompt taxonomy and complete prompt list
• Brand and competitor inclusion rules
• Raw answer captures
• Source and citation classification rules
• Repeat-run policy
• Missing-data and invalid-answer rules
• Scoring formulas
• Quality-control checks
• Limitations and version history

Until those materials exist, the responsible conclusion is not that one retailer or brand leads the market. It is that retail AI visibility should be measured as a documented pattern of mentions, recommendations, citations, and competitor appearances across defined shopper questions.

That standard gives decision-makers something more valuable than an unsupported ranking: a repeatable way to understand where their brands are visible, where competitors are present, and which sources shape AI-generated shopping answers.