Do Business Directories Still Matter? What the AI Citation Data Says

A declaration of interest

I want this to be true, so weigh what follows with that in mind. I run Local Media Ltd, which operates niche directories including Justice Directory, and I set out the case for owning one in The directory moat. I also run Local SEO Ltd, which sells AI visibility work. Most of the publishers quoted below have a similar stake, and their interests are listed in the table near the end.

Key figures at a glance

People now ask AI for a local business

BrightLocal's Local Consumer Review Survey (February 2026, 1,002 US adults) reports that 45% of consumers used AI tools for local recommendations, which made them the third most popular source of business recommendations. BrightLocal does not publish the question wording and the panel is American, so treat this as a direction of travel for UK businesses and not as a UK measurement.

Being named is also scarce. SOCi's 2026 Local Visibility Index (January 2026, nearly 350,000 locations across 2,751 multi-location brands), reported by Search Engine Land, found ChatGPT recommended 1.2% of locations, Perplexity 7.4% and Gemini 11.0%, against 35.9% appearing in Google's local 3-Pack. Those are chains, not independents, so the percentages will not transfer neatly to a small firm. The gap between being findable and being named is still the useful part.

That raises a practical question: what are these assistants reading before they answer?

How an assistant builds a local shortlist

With web search switched on, an assistant runs searches, reads what comes back and names a handful of businesses. In 8ight's study (October 2026; 1,630 answers through the API, covering 39 trades in 49 US and Greater Toronto places), the median answer named at least five businesses on ChatGPT and seven on Claude. It publishes the queries behind its figures, which is why I rely on it heavily. 8ight sells AI visibility services and says so, and its Claude sample is small at 218 answers.

One distinction matters throughout this article. A citation is a source shown with an answer. Training is what a model absorbed earlier, and nobody outside the AI companies can inspect it. Everything below measures citations, so it tells you where engines read at the moment of answering. It does not tell you which source changed the answer.

Different engines also use different routes to local data. Local Falcon's explainer (November 2025) lists Bing Places, business websites, directories and review sites among the places ChatGPT draws on. It offers no test data, so read it as informed opinion from a vendor. It is still a fair reminder that Google Business Profile is not the only listing worth maintaining.

Directories are among the most-cited sources

Yext analysed 6.8 million citations from 1.6 million queries per model on ChatGPT, Gemini and Perplexity (published October 2025, data from July and August 2025). Websites took 44%, listings 42%, reviews and social 8%, and forums 2%. Yext sells listings software and the sample came from its own client and prospect locations, so this is a friendly dataset. Even so, listings sit a close second.

BrightLocal's directory study (August 2026; 60,970 checks for 1,355 locations in the US, UK and Australia on AI Overviews, AI Mode and ChatGPT) recorded 1,967,762 citations. Google domains led with 28.63%, which partly reflects Google citing itself in its own products. Yelp followed with 9.53%. Older names barely register by share, at under 0.5% each for MapQuest, BBB and Angi, yet they were cited for many businesses: 54.69%, 29.30% and 18.97% of locations respectively. A low share of citations can still mean wide reach.

8ight's numbers point the same way. A directory or review site appeared in 99.5% of Claude's local answers and 65.3% of ChatGPT's.

Each engine reads different directories

This is where a single "get listed" recommendation stops working. In 8ight's data, Claude cited Yelp in 202 of 218 local answers (92.7%), while ChatGPT cited it in none of 1,412. ChatGPT leaned on the Better Business Bureau instead (38.7%, against 6.4% on Claude).

BrightLocal's study then contradicts that on the very same platform. It found Yelp to be ChatGPT's most common local source, appearing in 80% of ChatGPT answers. The two were published within weeks of each other. 8ight tested trades through the API in North America, while BrightLocal covered a wider mix of businesses in three countries, including restaurants. Those differences may explain the gap, but neither study set out to test why, so I would not state a general rule about ChatGPT and Yelp from either one.

Other studies add to the picture without settling it:

• Yext's model split found OpenAI models favoured listings (48.73% of citations, with more than 465,000 from Google), Gemini favoured websites (52.15%), and Perplexity drew heavily on MapQuest and TripAdvisor.
• Foundation Marketing with AirOps (28 million AI responses; 2026) counted 512,680 Yelp citations in Q4 2025, 3.4 times BBB's total. That comparison covers six chosen platforms and excludes Google Maps and TripAdvisor, and both publishers sell AI visibility services.

The sensible reading is to cover the several platforms each engine might use, and then to check your own sector, because a result from one engine or one country tells you little about the next.

Specialist directories carry more weight in some sectors

Yext's split by sector shows how far the mix moves. In healthcare, listings made up 52.6% of citations. In financial services, brand websites led with 48.2%, and in retail first-party websites led with 47.6%.

Legal

Legal has attracted the most research, nearly all of it American and most of it from agencies or directory owners.

• 5W's Legal AI Visibility Report (updated October 2026; 288 answers across four engines, measured on one day) found state bars named most often at 57.3%, followed by Martindale 47.6%, Avvo 46.2%, Super Lawyers 35.4% and Justia 35.1%. Chambers reached 23.3% and Legal 500 11.1%. 5W is a PR agency that sells this kind of work.
• Forward Push (July 2026; 38,000 searches in 15 US cities) found that 96% of Perplexity's legal answers cited at least one legal directory, while 68% cited the firm's own website. Forward Push markets law firms, and these citation figures cover Perplexity alone.
• Custom Legal Marketing (July 2026; 3,200 prompts in each of ChatGPT, Claude and AI Mode; 19,692 firms) found only 578 firms, or 2.9%, were named by all three. ChatGPT and Claude overlapped 37.5% on directory citations, more than four times their agreement on firms. Each prompt ran once, so some of that disagreement may be ordinary run-to-run variation.
• A white paper from Internet Brands, owner of Martindale and Avvo, says firms on several directories rank higher and appear more in AI answers. Its 177-firm sample covers Google rankings only, it gives no effect size, and it promotes the directories it studies, so I give it little weight.

Engines seem to agree more on which directories to read than on which firms to name, which is the most interesting legal finding here. It suggests a listing puts a firm in the material being read, with no promise of what happens next.

The UK picture

UK evidence is thin. Whito's study (July 2026; 36 questions, 18 each on ChatGPT and Gemini, across six categories and three towns) found Reddit was ChatGPT's most-cited source, in 10 of 18 answers, while Yell, Yelp and Facebook were cited zero times. Gemini gave no sources in 13 of its 18 answers, and otherwise cited only Checkatrade, TrustATrader, MyBuilder and the Gas Safe Register. Whito says it sells research files, not marketing services. The sample is far too small to generalise from, but it does caution against copying a US directory list.

The sensible UK equivalents depend on the trade. For law firms that means the SRA's Solicitors Register, the Law Society's Find a Solicitor, Chambers, Legal 500 and specialist directories. For trades it means the Gas Safe Register, NICEIC and the recognised review platforms. That list is my judgement from the US pattern and Whito's small sample, because no study above measures UK legal directories, including Justice Directory.

Directories alone are not enough

Four other signals appear alongside directories in the data, and the type of directory seems to matter as well.

Reviews. In 8ight's data (2,775 businesses with review counts), ChatGPT named 3.4% of businesses with fewer than 25 Google reviews and 22.2% of those with 1,000 or more. The median star rating was 4.9 for named and unnamed businesses alike, so volume was associated with being named while rating was not. SOCi did report a 4.3-star average for locations ChatGPT recommended, but that is an average across chains.

Your own website. Websites were the largest source in Yext's data. In 8ight's data, ChatGPT cited a business's own site more often than a directory (68.1% against 65.3%), though Claude did the opposite (45.0% against 99.5%).

Accuracy. Insites (2026; 10,000 US businesses) found about 93% were known to ChatGPT and Perplexity, but only 55.58% of ChatGPT results and 48.05% of Perplexity's fully matched the business's Google Business Profile. That is a strict test, since one wrong detail counts as a miss. Inconsistent listings are one plausible cause, and tidying them is cheap. Insites sells AI visibility tools to agencies.

Public records. ChatGPT cited a government or licensing site in 25.1% of 8ight's answers, and about 7 in 10 of those citations went to licensing boards, licence look-ups or contractor registers. Claude's figure was 0%. Whito's separate four-query study of Google's AI Mode also found it checked an accreditation register each time. Keeping regulator entries current costs little.

The type of directory also seems to matter. Whitespark's 2026 Local Search Ranking Factors (November 2025; 47 experts scoring 187 factors) ranked expert-curated "best of" lists first for AI visibility. Structured citation quality came 26th and citation quantity 40th. This is an opinion survey from a company that sells citation services, which makes its low ranking for bulk listings more striking. Insites similarly names reviews, profile completeness and Local Pack presence as its strongest correlates, and mentions directory coverage without a figure.

What the evidence does not show

• Cause and effect. None of these studies adds a listing and measures what changes. They show where engines read, and 8ight states plainly that it did not test cause.
• Stable results. 8ight asked ChatGPT the same question twice, a median of 39 minutes apart, and only 29% of the businesses named appeared in both answers (267 pairs, from one region and a few trades). ChatGPT and Claude both named just 28% of the 294 businesses recommended by either.
• Agreement between studies. BrightLocal's 2024 study (800 manual searches, 20 US cities) put directories at 15% of ChatGPT's local sources, against 58% for websites, with Yelp absent. It ran weeks after ChatGPT Search launched and BrightLocal's own later work looks very different, but it shows how quickly these findings can date. Reddit swings too: 6 of 1,630 answers in 8ight's data, 10 of 18 in Whito's.
• Independence. Almost every study comes from a company selling listings, software or AI visibility services. This article comes from someone who owns a directory.
• UK relevance. Apart from Whito and the UK share of BrightLocal's sample, all of this data is from North America. API results may also differ from what a person sees in the app, as 8ight and Yext both note.

A practical checklist

These steps are my reading of the evidence above. They are low-cost and worth doing for customers and search in any case, with no guarantee of an AI recommendation.

• Complete and maintain the core profiles: Google Business Profile, Bing Places and Apple Business Connect, which BrightLocal notes as a ChatGPT source. Add Yelp where your customers use it.
• Identify the two or three specialist directories and "best of" lists that matter in your sector. Favour selective lists that assess businesses over mass submission.
• Keep your name, address, phone number and services identical everywhere, starting with your main profiles.
• Keep regulator and licensing entries current if you work in a regulated trade or profession.
• Build a steady flow of genuine reviews. Review volume showed a clearer link than star rating.
• Keep your website clear and complete, because it is still a leading source. My AI Readiness Framework covers what to check.
• Test each engine separately and repeatedly, using the questions your customers would ask. One screenshot tells you very little when the list changes between runs.

Frequently asked questions

Do business directories help with AI visibility?

They appear to help AI assistants find and read about a business, as directories and listings are among the most-cited local sources in several 2025 and 2026 studies. No study shows that adding a listing causes a recommendation.

Which directories matter most for AI search?

It varies by engine and sector. Google, Yelp and BBB feature in US-led studies, but the studies conflict on Yelp and ChatGPT. In regulated sectors, specialist directories and official registers look more relevant.

Should I pay for mass directory submission?

The evidence gives little reason to prioritise it. Experts in the Whitespark 2026 survey ranked curated best-of lists first for AI visibility and structured citation quantity 40th. It is an opinion survey, but selective listings look the better use of time.

Do these findings apply to UK businesses?

Only in part. Nearly all the data comes from North America, and the one UK-only study here had just 36 questions. Use it as a guide to what to test, and check UK platforms and regulators for your own sector.

How can I check whether AI assistants recommend my business?

Ask each engine the questions your customers would ask, and repeat them several times. In the 8ight study, only 29% of businesses named reappeared when the same ChatGPT question was asked again, so a single result proves little.

The studies used

Every figure above was checked against the publisher's own page on 4 November 2026, except SOCi's, whose Perplexity and Gemini accuracy figures I could only see reported by Search Engine Land, so I have left them out.

Sources

• 8ight, Ask an AI for a roofer: https://8ight.ai/research/how-ai-recommends-local-businesses/
• BrightLocal, AI directory sources: https://www.brightlocal.com/resources/ai-directory-sources/
• BrightLocal, Local Consumer Review Survey: https://www.brightlocal.com/research/local-consumer-review-survey/
• BrightLocal, Uncovering ChatGPT Search Sources: https://www.brightlocal.com/research/uncovering-chatgpt-search-sources/
• Yext, AI citations: https://yext.com/about/news-media/ai-citations-release
• sector release: https://investors.yext.com/news-events/press-releases/detail/376/yext-research-86-of-ai-citations-come-from-brand-managed
• Foundation Marketing with AirOps, Yelp: https://foundationinc.co/lab/yelp-ai-local-discovery/
• SOCi, Local Visibility Index: https://www.soci.ai/insights/lvi/
• Insites, AI visibility study: https://insites.com/blog/what-ai-uses-to-choose-local-businesses-and-what-it-ignores/
• Whitespark, Local Search Ranking Factors: https://whitespark.ca/local-search-ranking-factors/
• 5W, Legal AI Visibility Report: https://www.5wpr.com/research/Legal-AI-Visibility-Report-2026
• Forward Push, law firm study: https://www.einpresswire.com/article/930058142?promo=4400
• Custom Legal Marketing, law firm overlap: https://custom.legal/knowledge/chatgpt-claude-and-google-ai-mode-rarely-recommend-the-same-law-firms/
• Internet Brands, legal directories white paper: https://www.martindale-avvo.com/wp-content/uploads/2026/05/WHY-LAW-FIRMS-IN-LEGAL-DIRECTORIES-DOMINATE-IN-GOOGLE-AI-SEARCH-2.pdf
• Whito, UK AI citation sources: https://whito.co.uk/research/ai-citation-sources-uk-study/
• Local Falcon, ChatGPT data sources: https://www.localfalcon.com/blog/chatgpt-local-search-data-sources-where-does-business-info-come-from
• Yext, AI citations by model: https://www.yext.com/research/article/ai-citations-user-locations-query-context
• Search Engine Land on the SOCi index: https://searchengineland.com/ai-local-visibility-report-2026-468085
• Insites, AI visibility benchmark: https://insites.com/blog/the-ai-visibility-benchmark-how-10000-local-businesses-actually-perform-in-chatgpt-and-perplexity/
• SRA Solicitors Register: https://www.sra.org.uk/consumers/register/
• Law Society, Find a Solicitor: https://solicitors.lawsociety.org.uk/
• Gas Safe Register: https://www.gassaferegister.co.uk/
• NICEIC, find a tradesperson: https://niceic.com/find-a-tradesperson/