Local Citations in the AI Era: Which Directories ChatGPT and AI Overviews Actually Pull From (2026)

Key takeaways

• Your own website still wins the most local AI citations. Steady Demand's analysis of 14,000+ local Gemini and ChatGPT citations found it takes 50% of them, more than every directory, forum and review platform combined.
• For ChatGPT, Yelp is the dominant directory by a wide margin. BrightLocal's study of 1.9 million citations lists 99,281 Yelp citations, ahead of Tripadvisor (10,227), OpenTable (9,147) and BBB (4,247).
• The right directories depend on the vertical and the engine. Lawyers, dentists and auto shops get cited from different source ecosystems, and ChatGPT, Gemini and AI Overviews do not overlap much.
• Rankings no longer predict citations. Ahrefs found only about 38% of AI Overview citations still rank in the top 10, down from about 76% in mid-2025.
• Reddit has dropped out of both ChatGPT and Gemini local answers, and directories absorbed the share. Treat all of these numbers as a snapshot and re-check them.

Why local citations got a second job

For about fifteen years, a local citation meant one thing: your name, address and phone number (NAP) listed consistently across directories, which nudged Google's local pack. That job still exists. But directories now do a second one. They are the pages AI assistants read when they decide which plumber, dentist or ramen shop to name.

That changes how you should prioritise. Under the old model you chased volume, 50 or 100 listings, because consistency was the signal. Under the new one, a small number of directories carry most of the weight for a given engine and a given vertical, and the rest are mostly noise.

One caveat before the data. The studies below come from third parties (a local SEO software company and a Google Business Profile agency), each with its own methodology and sample. Promptwatch Data has no report on local directories specifically, so where I use it, I use it for adjacent patterns and say so.

What the big local studies found

BrightLocal: 1.9 million citations across 116,670 domains

BrightLocal analysed citations from Google AI Overviews, Google AI Mode and ChatGPT for local queries. Its sample leans toward restaurants and bars (about 20% of locations), with home and trade services at roughly 11%, financial and professional services at about 9.7%, and retail at about 4.9%.

The ChatGPT table is the interesting one. ChatGPT looked at the fewest sites of the three platforms, and Yelp sits far above the rest:

Two things worth flagging. First, I could not verify the publication date of this study, so check it before you quote it as current 2026 data. Second, the page also contains a line saying Yelp appeared in 80% of ChatGPT answers but was "cited only 600 times," which conflicts with the table's 99,281 figure. I would trust the table for relative ranking and not repeat the 600 number.

Across all three platforms, Yelp makes up 8.53% of all citations, and BrightLocal calls it the second most-used source for AI Overviews and AI Mode but the most-used for ChatGPT. The skew toward restaurants probably inflates the OpenTable and Tripadvisor numbers. If you run a law firm, do not assume that table is your table.

Steady Demand: 14,000+ Gemini and ChatGPT local citations

Steady Demand's AI Citation Ledger is the more useful read if you want to understand how messy this is. It lists nine findings, and several matter directly for citation strategy.

• The business's own website wins 50% of all citations. Directories matter, but they are not the main event.
• AI-recommended businesses average 4.76 stars, but a plain Google search baseline averages 4.82. The quality bar was already there before AI.
• There is no universal playbook. Lawyers, dentists and auto shops are cited from completely different source ecosystems.
• A handful of national platforms and brands appear in every metro. That is the floor, not the differentiator.
• The long tail still belongs to individual businesses.
• Ask Gemini the exact same question twice and the cited sources match only about 41% of the time. Steady Demand calls this Grounding Drift.

That last one should make you cautious about any single citation snapshot, including the tables in this article. If the same question produces different sources on back-to-back runs, one scan proves very little. You need repeated measurement over weeks.

The Reddit reversal

The Steady Demand update dated Sept 9, 2026 reports that Reddit fell from 13.7% to 1.7% of Gemini's local citations, about five weeks after the same thing happened in ChatGPT. Directories absorbed the share in both. In July, one forum had outcited Angi, Thumbtack and HomeAdvisor combined. By September it trailed them 1.7% to 19.1%.

Promptwatch Data shows the same pattern on the ChatGPT side: reddit.com's share of ChatGPT citations fell from about 4% to 0.5% on August 14, 2026, with AI Overviews and AI Mode declining more gradually. The practical read for local businesses is simple. If you spent 2025 planning a Reddit presence for local discovery, the engines have moved on, and listings on structured directories matter more again.

What Promptwatch Data adds (and where it does not reach)

Promptwatch has no report on local directory citations or local-pack results, so none of what follows is local-specific. Three reports are still useful for understanding the shape of the problem.

Branded prompts lean on your own pages and on reviews

In Promptwatch's ChatGPT content-type data (Sep 28 to Oct 4, 2026), product pages are 22.4% of classified citations and landing pages 16.8%. For branded prompts, meaning ones that name a business, product pages are 23.3% and landing pages 22.8%, and reviews rise to 5.6%, more than three times their share across all prompts.

That lines up with Steady Demand's 50% own-site finding. When someone asks ChatGPT about a specific business by name, it leans on that business's own pages first and review sites second.

Verticals with directories: travel and automotive

Travel and automotive are the closest Promptwatch reports to local directory behaviour.

For travel prompts (Sep 28 to Oct 4, 2026), Skyscanner has 0.78% of citations, Going 0.52% and Tripadvisor 0.49%. Booking and search sites earn 2.42% of all travel citations, and review and advice sites like Tripadvisor and Trustpilot earn 0.86%.

For automotive prompts (August 2026), Autotrader has 5% and Cars.com 4.6%, ahead of every manufacturer. Chevrolet.com is the top manufacturer site at 2.62%. Marketplaces and listings have 14.65% and review sites 12.62%. One UK dealer group's own site reached 0.28%, so a local business page can sit beside national sites.

The lesson is the same across both. In directory-heavy verticals, the platform that owns the transaction or the review beats the brand. Keep your profile complete on those platforms.

Healthcare is its own world

For healthcare prompts, ChatGPT cites NIH, the FDA and the NHS most, and government and public health sites earn almost half of citations (Sep 9 to Oct 8, 2026). A local clinic cannot win those slots, which is why Steady Demand finds dentists being cited from a different ecosystem than, say, auto shops. For a practice, your own site and the major review and provider directories matter more than trying to compete with the NIH.

How the engines differ

The engines pull from different places, and you should not treat "AI" as one channel.

The Ahrefs finding is also worth keeping in mind. Ahrefs analysed four million AI Overview URLs and found only about 37.9% of cited URLs also rank in the top 10 organic results, down from about 76% in mid-2025. In a separate study of 15,000 prompts, Ahrefs found just 12% of URLs cited by assistants like ChatGPT, Gemini and Copilot rank in Google's top 10 for the original query. Being on page one of Google is no longer a reliable proxy for being named in an AI answer.

Where on the page citations come from

One more finding that matters for your own location pages, since they are your biggest source of citations. CXL's 100-page study of AI Overview citations found 55% of cited passages appeared in the first 30% of the source page. Kevin Indig found a similar pattern across 18,012 ChatGPT citations, where 44.2% came from the first 30% of a document.

CXL is careful to call this a retrieval consideration, not a guaranteed ranking factor. For a local page, though, it is cheap to act on. Put the business name, service area, what you do and a clear answer to the likely question (hours, pricing range, who you serve) near the top, not buried under a story about your founder.

A practical directory priority list

Given the data, I would work in this order.

1. Your own website. Build real, specific location and service pages, with the key facts early in the page. This is where half of local citations land.
2. Google Business Profile. Gemini draws on live Google Places lookups for businesses it names, and AI Overviews sit on top of Google's own data. I did not research GBP specifics for this piece, so treat this as a baseline you already have covered.
3. Yelp. It is the biggest single directory source in the BrightLocal data, and the top one for ChatGPT. Claim it, complete it, and keep the NAP consistent.
4. Vertical platforms. Tripadvisor and OpenTable for hospitality, Autotrader and Cars.com for automotive, Angi and HomeAdvisor for trades, plus the review sites that apply to your category.
5. BBB and Trustpilot. Both show up in the ChatGPT table. They are not glamorous, but BBB's 4,247 citations across 354 locations says it is read.
6. Apple. Apple.com appears in the ChatGPT table with 600 citations across 134 locations. Keep your Apple Maps and Business Connect data current. I did not research Bing Places or Apple specifics beyond this.
7. The long tail of generic directories (Yellow Pages and similar). Keep them accurate if they are cheap to maintain, but do not spend a week on them.

One more practical point: mid-authority sites are not locked out. Promptwatch Data on ChatGPT citations by domain rank shows sites with domain rank 46-75 earned almost half of ChatGPT citations in August 2026, while DR 91-100 fell to about 3%. That is general, not local, but it suggests small and mid-sized directories and niche vertical sites can matter.

Tools for managing listings and measuring AI visibility

There are two jobs here. One is keeping listings accurate. The other is finding out whether AI is citing them.

Listing and reputation management

These tools handle the traditional citation side: pushing consistent data to directories, monitoring reviews, and managing multiple locations.

BrightLocal now includes AI visibility reporting alongside rankings, listings and reviews. Yext, Chatmeter, Reputation.com, SOCi and Synup lean toward multi-location brands, where managing hundreds of listings by hand is not realistic. For a single-location business, a manual pass through the top five directories is often enough.

Measuring what AI cites

Listings work is only half the job. Given Grounding Drift, where identical Gemini prompts return matching sources about 41% of the time, you need ongoing tracking, not one screenshot.

Promptwatch tracks ChatGPT, Google AI Overviews, Google AI Mode, Gemini and other engines through their real interfaces, with country tracking on every paid plan and state/city tracking and exact-coordinate local targeting on Business. That lets you see which directories and pages AI cites for your location queries, then check whether fixes changed anything.

Birdeye Search AI is built specifically around location-level AI visibility, which fits multi-location brands.

Some other tracking tools in this space, depending on budget and scale:

For more options, the GEO software directory at bestgeosoftware.com and the AI rank tracking directory at ai-rank-tools.com list a broader set.

How to run your own local citation audit

You do not need a tool to start. This takes an afternoon.

1. Write 15 to 20 prompts a real customer would type. Include "best [service] in [city]", "[service] near me", and a few branded ones with your business name.
2. Run each in ChatGPT, Gemini, and Google (watching for AI Overviews and AI Mode). Run them more than once, on different days, because of the drift problem.
3. Log every source named or cited. Group them: your site, Yelp, other directories, review sites, news, social.
4. Note which directories show up repeatedly in your vertical and city. Those are your priority list, not the generic top 10.
5. Check whether your listing on each one exists, is claimed, and has accurate NAP, hours, categories and photos.
6. Compare your rating and review count with the competitors AI names. Steady Demand found AI-recommended businesses average 4.76 stars, so low ratings are a real handicap.
7. Repeat monthly.

What I would not do

I would not buy a 100-directory submission package. The data points the other way: concentration around a handful of sources per engine and vertical.

I would not build a local strategy around Reddit. It collapsed in both major engines this year, and it may move again.

I would not trust any single number here as permanent. BrightLocal's date is unverified, Steady Demand's own report was updated in September after engines changed, and Promptwatch Data periods are one-week snapshots in some cases. Treat them as a direction, then measure your own market.

If you want help turning citation and AI visibility data into a local SEO and GEO plan, 1001 SEO Media runs this work as a service, and the Promptwatch Data hub at promptwatch.com/data is the place to track how engine behaviour keeps shifting.

Bottom line

In 2026, local AI citations come first from your own website, then from a short list of directories that varies by engine and vertical. Yelp leads for ChatGPT. Vertical platforms matter in travel, automotive and trades. Reddit is out for now. Fix your own pages, claim the five or six directories your audit surfaces, and measure repeatedly, because the same prompt will not return the same sources twice.