First on Google does not guarantee a mention in ChatGPT

The comparison applies when Google's first result was classified as a supplier's own website. It measures whether the business was named in the answer text.

For a business that has worked hard to reach the top of Google, that is a reason to check what happens when the same buying question moves to an AI service. A strong Google position alone does not establish whether the business will appear there.

The study used a fixed set of 1,000 questions across 25 commercial categories, covering software, services and ecommerce. Each question was submitted with identical wording to Google and six consumer AI services: ChatGPT, Microsoft Copilot, Gemini, Perplexity, Google AI Mode and Google AI Overviews.

Every question was tested in three rounds during 6 to 7 September 2026. Google was searched three times too, providing a matching comparison for each round.

The collection produced 17,537 completed AI responses and 3,000 Google captures. Discoverable collected the answers from consumer websites using browser automation. The study did not generate its test answers through model APIs; AI models were used in parts of the subsequent classification and checking work.

The questions and protocol were fixed before collection. The supporting evidence includes the question set, calculation tables, analysis code, correction notes and saved records for the three examples below.

Across all six services, the business in Google's first organic position went unnamed in 66.2% of scored comparisons.

The result differed by service. It also depended on whether a business linked as a source, without being named in the answer, counted as present.

Perplexity shows why the distinction matters: its figure falls from 74.9% to 55.9% when citations count. A business can be used as a source without being named in the answer itself. These results describe overlap with Google, rather than ranking the quality of the AI services.

The pattern also extended to Google's first three organic positions. Across all six services, eligible businesses occupying those positions went unnamed in 71.3% of individual business and answer comparisons. That does not mean all three businesses were absent together in 71.3% of answers.

All three businesses below ranked first on Google in every round for the stated question. The examples were selected after collection to illustrate different outcomes; they are not a representative sample.

Buying coffee for a moka pot. For the question “cheap coffee for a moka pot”, Iron & Fire ranked first in all three Google captures. It was not named in any of the 17 available AI answers, and none of the captured citations pointed to its website. One AI Overview did not appear and was excluded.

ChatGPT's first answer suggested options including Lavazza. Its third asked for the shopper's location. A question about cheap coffee can reasonably produce supermarket suggestions, so this example does not establish that Iron & Fire should have been recommended. It shows how different the business's exposure was for the same question.

Setting up a first company. For “company formation for a first time founder”, Your Company Formations ranked first in all three Google captures. None of the 18 AI answer texts named the business, but Perplexity cited its website in every round.

The ChatGPT answers mainly explained how to set up a company. That is useful context: an answer can help someone with the process without giving them a list of formation agents. Calling this business completely invisible to AI would also overlook its citations.

Selling digital products. For “free platform for selling digital products”, Payhip ranked first on Google in all three rounds and was named in all 18 AI answers.

Google visibility and AI mentions can align. The study does not establish why Payhip appeared consistently, or whether it would do so for other questions or on another day.

Google presents a list of links. An AI service may offer a shorter selection, explain a process or ask a follow-up question. An answer that gives advice without naming suppliers can still be counted in this study.

To explore that difference, we also restricted the ChatGPT comparison to answers that named at least one supplier from the matching Google top ten. In that narrower set, the number-one business went unnamed in 54.0% of answers. This filter only recognises suppliers in the Google comparison set; it does not identify every answer that mentioned any brand.

The research did not measure lost sales, decide which supplier deserved a recommendation or establish why particular businesses appeared. It measured whether Google-visible businesses were named in the answers collected.

Discoverable designed the study, collected the data and carried out the analysis. We extracted organic positions, page titles, links and snippets from rendered Google results pages. Ads, maps, People Also Ask and the AI Overview were excluded from the organic comparison. Google's organic results and its AI Overview came from the same collection call.

The comparison used websites classified as suppliers' own sites. Publishers, directories, marketplaces and retailers were excluded under the study's rules. “Number one” means actual organic position one, rather than the first supplier remaining after other results were removed. Multiple pages belonging to one business counted once, at its best position.

AI answers were checked for recorded business names, verified aliases, domain forms and attributed product names. Specified spelling and separator variations were allowed. Unrecorded trading names could still be missed.

All requests specified country GB. The records do not independently establish the actual search location or whether sessions were fresh and logged out. Most underlying model versions were unavailable, so these are observations of consumer services rather than a controlled comparison of selected model tiers.

The three rounds took place within one collection window. The question set was designed for this study and was not weighted by real-world search volumes. The findings should not be treated as estimates for all searches or all countries.

A useful starting point is to take the buying questions that matter to the business and check both the Google results and the AI answers. Record whether the business is named, whether its website is cited, which alternatives appear and whether the answer offers suppliers at all.

Repeat the checks before drawing conclusions from a single answer. The aim is to understand the business's visibility for specific questions, without assuming that its Google ranking tells the whole story. The free AI visibility check runs that comparison for one domain.

The evidence pack includes the exact questions, results tables, methodology, analysis code and saved records for the three examples. It contains selected raw records; reproducing the full collection requires the retained study repository.

Suggested citation: Discoverable, UK AI Shortlist Study, September 2026. Data collected 6 to 7 September 2026; analysis dated 14 September 2026. Please link to this study page when reporting the findings.

Media enquiries: Jon David, Media Relations, Discoverable, press@bediscoverable.ai.

Discoverable is an AI visibility software company and has a commercial interest in this subject. It helps businesses understand where they are named and cited in AI answers and plan improvements.