--- title: "AI visibility: how to measure it (2026 data) | Aristral" description: "AI visibility data from 60 UK searches and 20 Bristol prompts shows who AI cites and recommends, with steps to check your brand across search." url: "https://aristral.com/blog/ai-visibility" canonical: "https://aristral.com/blog/ai-visibility" --- [Home](https://aristral.com/)/[Blog](https://aristral.com/blog)/[SEO](https://aristral.com/blog/category/seo) SEO # How to Measure AI Search Visibility AI visibility measures whether AI search mentions, cites or recommends your business. Our 60 UK searches and 20 Bristol prompts compare answers with Google results. TBTaha Bilal·2026-09-25·9 min read ![AI visibility illustrated as a search bar feeding bright answer cards and cited source tiles into a chat window](https://aristral.com/_next/image?url=%2FBlogImgae%2Fai-visibility-hero.webp&w=3840&q=75) ## What does presence in AI search measure? AI visibility measures how often an AI search answer mentions, cites or recommends a business for relevant questions. Aristral tested 60 everyday UK searches on 25 September 2026: 51 showed an AI Overview, including 49 of 50 non-local searches. Only 48% of the cited URLs also appeared in the organic top 10 for the same query. In a separate test, 6 of 107 businesses recommended in 20 Bristol service prompts appeared in the Google Maps top three for that service. The Google sample measured citations; the Bristol test compared recommendations with Maps. They record different outcomes. Track a fixed set of customer questions, save the answer and cited pages, then repeat the same prompts across engines and dates. Treat each score as a sample tied to its prompts, location, engine and run count. To measure AI search visibility, run a fixed set of customer questions across the same engines and locations, then record brand mentions, page citations and recommendations separately. Save the answer and cited URLs, and compare Google AI Overviews with organic results. Google Maps is a separate measurement. ## How AI search visibility works For a fixed set of customer searches, count how often an AI answer names the business or its website. A mention names the business. A citation links to a page. A recommendation puts the business forward as an option. Record each separately: a name without a link tells you something different from a cited service page or a direct recommendation. A brand may appear in an answer without a link to its site. A page may be cited without the business being recommended. Save the wording, citation URL and recommendation for each prompt, then count each result against the same prompt set. We treat this as a measurement question rather than one universal score. Our [GEO vs SEO guide](https://aristral.com/blog/geo-vs-seo) explains how search optimisation differs from work aimed at generated answers. ## Where these answers appear Check each search product separately. Google AI Overviews, Google AI Mode, ChatGPT search, Perplexity and Copilot can answer the same question differently. Keep the prompt, location and date with each result. Google includes AI Overview and AI Mode activity in Search Console's Web search data. Its Generative AI features report gives a separate view in the interface. See Google's [AI features guidance](https://developers.google.com/search/docs/appearance/ai-features) and [report announcement](https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports) for details. | Surface | What to record | Useful comparison | | ------------------- | ------------------------------------------------------- | ------------------------------------------------ | | Google AI Overviews | Answer text, cited URLs, query and date | Organic top 10 for the same query | | Google AI Mode | Mention, linked sources and answer position | The matching Search Console generative AI report | | ChatGPT search | Business mentions, citations and recommendation wording | The same prompt in another engine | | Perplexity | Answer, cited URLs and date | Repeat runs on the same fixed prompt | | Copilot | Brand mentions and linked sources | The same question in search and other assistants | The Search Console Generative AI features report shows impressions for AI Overviews and AI Mode. It has no queries tab or API, and those impressions are also included in the usual Web search totals. Use SERP data and saved prompt runs for per-query checks. For other engines, keep a prompt log and review referral analytics. Microsoft Clarity documents AI bot and visibility features; check its [AI Visibility overview](https://clarity.microsoft.com/ai-visibility) before relying on a feature in your account. ## Test 1: AI Overviews on 60 UK searches AI Overviews appeared on 51 of the 60 searches. The query group made a clear difference. Forty-nine of 50 non-local searches showed an Overview. Two of ten service-plus-town searches did, while nine showed a local pack. We pulled the results through the DataForSEO SERP API on 25 September 2026. | Query group | Overviews | Avg sources | Cited URLs in top 10 | | -------------------------- | --------- | ----------- | -------------------- | | B2B marketing and software | 10 of 10 | 4.5 | 53% | | Beauty and aesthetics | 10 of 10 | 7.2 | 39% | | Accounting and tax | 10 of 10 | 8.4 | 39% | | Home improvement | 10 of 10 | 7.4 | 57% | | Legal | 9 of 10 | 5.3 | 56% | | Service plus town | 2 of 10 | 2.5 | 80% | | All groups | 51 of 60 | 6.4 | 48% of URLs | AI Overview presence by query group Share of searches in each group that showed an Overview, calculated from ten queries per group. 100% 100% 100% 100% 90% 20% B2BBeautyAccountingHomeLegalService + town Source: Aristral DataForSEO SERP API snapshot, 25 September 2026\. Percentages calculated from the ten queries in each group. Across the 51 Overviews, we counted 328 citations from 235 domains. Only 48% of cited URLs also ranked in the organic top 10 for the same search. The domain overlap was 56%. The number one organic result's domain appeared in 29 of 51 Overviews. A first-place result did not guarantee a citation. To check AI Overviews citations, compare each linked URL with the organic top 10. YouTube was the most-cited domain, with 35 of 328 citations (10.7%). Government guidance and specialist sites appeared too. Check the source list for your own topic before assuming AI answers reuse only the highest-ranking pages. This was one day's UK desktop snapshot across 60 queries. It is not a trend or forecast. The local searches used service-plus-town queries. We did not test every city, device or search history. Repeat the same query set in your market to see whether the pattern holds. ## Test 2: ChatGPT recommendations vs Google Maps ChatGPT's Bristol recommendations seldom matched the Maps top three in our test. We sent 20 prompts, including "Which dentist in Bristol would you recommend? Give me the business names." The answers named 107 businesses. Six also appeared in the Maps top three for the same service. ChatGPT picks found in Google Maps Share of 107 recommended businesses that also appeared within each Maps rank range for the same service. 6% 7% 14% Maps top 3Maps top 10Maps top 20 Source: Aristral DataForSEO AI Optimization and Maps APIs, 25 September 2026. ChatGPT named six businesses among the 60 Maps top-three places across those searches. For ten services, none of its recommendations appeared in the Maps top 20\. We matched results by website domain or business name. The answer to the "SEO agency" prompt included Aristral once. That was one answer from one run, not a stable ranking. ![Two side-by-side answer panels recommending different shops on the same street, showing how AI visibility differs by engine](https://aristral.com/_next/image?url=%2FBlogImgae%2Fai-visibility-two-answers.webp&w=3840&q=75) The same local service question can produce different shortlists in AI answers and map results. We used gpt-4o-mini with web search through the DataForSEO AI Optimization API. This was not the consumer ChatGPT app. We ran each prompt once. Maps results came from one central Bristol coordinate, so businesses in outer districts may rank lower from that point. Treat the figures as directional. Our [Google Maps ranking guide](https://aristral.com/blog/google-maps-ranking) covers the local pack side. ## How to check your presence in AI search Start a spreadsheet before choosing an AI visibility checker. Keep prompt wording, date window and location the same each month. Save the answer or source links where possible. That makes changes easier to compare. 1. Write 20 to 30 fixed prompts from real customer questions. Include product, service, comparison and local intent where they fit your business. 2. Run the same wording in three AI search engines once a month. Record the engine, date, country, device and whether web search was enabled. 3. For each answer, mark whether the brand was mentioned, recommended or linked. Save every cited URL and note the order in which the answer presented options. 4. Check Google AI Overviews for your target searches and compare cited URLs with the organic results. Record when no Overview appears. 5. Use Search Console's Generative AI features report to review impressions from Google AI Overviews and AI Mode. The report is in the interface, has no queries tab or API, and its impressions are included in Web search totals. For query-level checks, compare SERP data with your prompt log. Google's [AI features guidance](https://developers.google.com/search/docs/appearance/ai-features) explains its Web reporting. 6. Review server logs or analytics for visits and crawler activity. Clarity may help with AI bot visibility when its features are available. Check its current documentation and settings before treating those signals as complete. A prompt log can include the question, engine, date, brand mention, recommendation, citation URL, answer position and notes. Count the prompts where you appeared and divide by the prompts you ran. Keep separate totals for mentions, citations and recommendations. Show the raw counts beside the percentage so readers can see what it represents. When you report AI visibility tracking, say whether you mean manual checks, automated prompt runs, Search Console reporting or referral measurement. Each method answers a different question. Our [llms.txt guide](https://aristral.com/blog/llms-txt) explains the file. Check crawler rules for each service you want to measure. ## What should a useful score tell you? A useful score has a clear method behind it. Ask which prompts, engines, locations and runs produced the number. Find out whether it counts brand mentions, source citations, recommendation position or a combined measure. Without those details, two scores may describe different results. | Method detail | What to ask | Why it changes the result | | ------------- | ------------------------------------------- | --------------------------------------------------------- | | Prompt set | Which exact questions were tested? | Different prompts surface different businesses and pages. | | Engines | Which AI search products are included? | Each engine can return different answers. | | Location | What country, city or user context was set? | Local recommendations depend on where the search is made. | | Runs | How many times did each prompt run? | One answer can change on a repeat run. | | Evidence | Can I export answer text and citation URLs? | Source records let you audit the score and act on it. | Use share of answers as a baseline: the number of relevant prompts where your business appeared divided by the prompts tested. Keep separate shares for citations and recommendations. Put the raw counts beside each percentage. Repeat the prompt set over time, and describe changes as observations rather than proof that one site edit caused them. ## AI visibility tools: what to look for Choose a tool whose measurement method you can inspect. Start with the output you need. A citation monitor may not track recommendations. A prompt tracker may not show whether a linked page ranks in organic search. Look for enough detail to check the results yourself. | Criterion | Check before choosing | | ----------------- | ----------------------------------------------------- | | Engines covered | Does it track the search products your customers use? | | Prompt control | Can you write and reuse your own customer questions? | | Location settings | Can you set a meaningful country or city? | | Repeat runs | Does it show how often each prompt was run? | | Citation export | Can you download source URLs and answer records? | | API access | Can the data join your reporting workflow? | Try a small prompt set before you settle on a tool. On the same day, compare its answer and citations with a manual run on the same engine. Ask how it handles missing answers, changed prompts and locations it cannot reproduce. If it only gives you a percentage, ask for the records behind it. ## How to improve AI search visibility Work on pages and sources that AI answers can use, then check whether they appear in your fixed prompt set. In our Google study, 48% of cited URLs also appeared in the organic top 10\. The rest came from URLs outside those ten results. Organic rankings still gave pages a route into the answers. 1. Write pages that answer the questions customers ask. Put the answer near the top, then add details and evidence specific to your business. 2. Check citations for your topic. Improve pages that already earn relevant impressions. Fill gaps where an answer relies on a source your customers expect you to explain. 3. Make the business facts consistent across your site and the profiles customers use. Keep service names, locations and contact details accurate and easy to confirm. 4. Publish useful material on channels found in your source list. YouTube accounted for 35 of 328 citations in our Google sample. That figure applies to these queries on this date. 5. Review access rules for the search crawlers you want to reach. Our [llms.txt explainer](https://aristral.com/blog/llms-txt) covers the file question; do not assume a text file overrides crawler access controls. 6. For local searches, improve your Google Business Profile and local citations alongside your website. In our Bristol test, AI recommendations and Maps results produced different shortlists. Check them separately. For a wider plan, see [AI Overviews optimisation](https://aristral.com/blog/ai-overviews-optimisation). Choose actions tied to pages and sources your checks actually find. Then repeat the same prompts and note which answer, citation or recommendation changed. ## How Aristral measures AI search At Aristral, we track Google AI Overview citations from SERP reference data and run fixed prompt sets through language model APIs alongside Search Console. We record brand mentions separately from linked citations and local recommendations. Those records help us decide which pages or listings to review. Our Bristol test found different shortlists in AI answers and Maps results. Each needs its own check. The comparison was one run per prompt, so it does not establish a stable ranking. If you need help with this work, see our [AI SEO service](https://aristral.com/agentic-seo) and [contact us](https://aristral.com/contact). ## Frequently asked questions What is AI visibility? It measures how often a brand or website appears in AI-generated answers to relevant prompts. Count mentions, linked citations and recommendations separately. A useful report names the prompts and engines tested, the search location and the number of runs. How do I check my AI visibility? Create a fixed list of customer questions, run them in several AI search products, and record brand mentions, recommendations and cited URLs. Repeat the checks on a schedule with the same location settings. For Google, use Search Console's Generative AI features report for impressions, then compare SERP citations with your query set. How do I judge a visibility score? There is no universal good score because providers may test different prompts, engines, locations and answer types. Ask for the method and raw observations. A share of answers is easier to interpret: relevant prompts where your business appeared divided by prompts tested, with the counts and sample shown. If I rank on Google, will ChatGPT recommend me? A high Google position does not guarantee a ChatGPT recommendation. In Aristral's one-day Bristol comparison, 6 of 107 recommended businesses appeared in the Google Maps top three for the same service. We used an API model with web search and ran each prompt once, so the result describes that sample only. Can Search Console show AI Overview traffic? Yes. Search Console includes AI Overview and AI Mode activity in Web search data. Its Generative AI features report gives a separate impressions view in the interface, with no queries tab or API. AI impressions are also included in regular Web search totals, so use SERP data and prompt logs for query-level checks. Do I need llms.txt to be visible in AI search? This study did not test whether llms.txt changes how AI answers use a site. Treat it as a separate file question, and first check whether the services you care about can reach your pages. Accurate business details and pages with useful evidence give you things to review. Read our [llms.txt guide](https://aristral.com/blog/llms-txt) for more on the file. ## ⚗ Methodology The AI Overview study used DataForSEO SERP API results for 60 English desktop searches in the United Kingdom on 25 September 2026, with ten queries in each of six groups. The Bristol comparison sent 20 service prompts to gpt-4o-mini with web search on, then compared recommendations with mobile Google Maps results around central Bristol. Each prompt ran once. This is a snapshot, not a trend. Google results and AI answers can vary by place, device, wording and time. We track AI Overview references from SERP data and run prompt sets through APIs alongside Search Console. Its Generative AI features report is UI-only, shows impressions, has no queries tab or API, and counts AI impressions within normal Web search totals. Check the query list, engine, location and current Search Console reporting before applying these findings to your own site. Technical review: Huzaifa Jan Asim, co-founder and CTO. Last reviewed 25 September 2026\. Corrections: admin@aristral.com About the author TB ### Taha Bilal Co-founder, Aristral Taha Bilal is a co-founder of [Aristral](https://aristral.com/), a UK AI automation and SEO agency based in Clifton, Bristol. He has been running SEO and digital-growth campaigns for SMB and SaaS clients since 2018, and now leads Aristral's combined SEO + GEO programmes for service businesses across the UK and US. Corrections and source requests: [admin@aristral.com](mailto:admin@aristral.com). [LinkedIn →](https://www.linkedin.com/in/tahabilal366/) On this page * [How AI search visibility works](#what-is-ai-visibility) * [Where these answers appear](#where-ai-visibility-appears) * [Test 1: AI Overviews on 60 UK searches](#overview-study-results) * [Test 2: ChatGPT recommendations vs Google Maps](#chatgpt-maps-comparison) * [How to check your presence in AI search](#check-ai-visibility-free) * [What should a useful score tell you?](#meaningful-visibility-score) * [AI visibility tools: what to look for](#choose-ai-visibility-tools) * [How to improve AI search visibility](#improve-ai-search-visibility) * [How Aristral measures AI search](#where-aristral-fits) Want this built for you? Book a free 30-minute call and we'll map the fastest path for your business. [Book a 30-minute call](https://aristral.com/book) Filed under: [SEO](https://aristral.com/blog/category/seo) ## Related articles [SEOllms.txt: what it is, who uses it and whether it helps8 min read](https://aristral.com/blog/llms-txt)[SEOAutomated SEO reporting: what we automate across 12 sites11 min read](https://aristral.com/blog/automated-seo-reporting)[SEOWhy Most "AI SEO Audits" Miss the Structural Issues That Actually Matter12 min read](https://aristral.com/blog/ai-seo-audit-structural-issues) --- Markdown version of this page. Human page: https://aristral.com/blog/ai-visibility. Site index for agents: https://aristral.com/llms.txt. Contact: https://aristral.com/contact