Does ChatGPT recommend the same home service businesses as Google?
Mostly not. Across 126 local home service questions in 21 U.S. markets, asked 3 times each through OpenAI’s API (gpt-5-mini with web search, standing in for ChatGPT) and Google’s Gemini API (Gemini 4 Hydrogen with Google Search) on Nov 13, 2026, 46% of the businesses ChatGPT named were also named by Gemini for the same question.
- Gemini named more: about 10 businesses per answer, against 5 for ChatGPT. Only 27% of Gemini’s picks were also in ChatGPT’s answers.
- Ask ChatGPT the same question 3 times and the list changes. 16% of the businesses it named showed up in all 3 answers, and 61% in just one. Gemini: 36% and 45%.
- Business websites were the biggest source: 72.3% of ChatGPT’s citations and 79.8% of Gemini’s. When either engine named a business with a known website, it cited that site most of the time (ChatGPT 69%, Gemini 88%). Neither linked to Yelp once.
- Of the businesses either engine named, ChatGPT named 72% of the chains (same name in 3+ of our markets) and 45% of the local ones. Gemini went the other way (66% vs 75%).
- Behind the answers, Gemini ran 8.6 Google searches per answer. 63% of them checked a specific business by name, and 26% were about price.
By Justin Hubbard, who runs Adimize and has owned and operated a junk removal and dumpster rental company for over 12 years
Based on 126 local questions × 3 runs × 2 engines (756 answers), 21 U.S. markets, Nov 13, 2026, via API
How often ChatGPT and Gemini agreed
| Measure | Share |
|---|---|
| Businesses ChatGPT named that Gemini also named (same question) | 46% |
| Businesses Gemini named that ChatGPT also named | 27% |
| Named by both, of every business either one named | 21% |
| One answer from each, same question: overlap | 19% |
Agreement by type of question
| Question | ChatGPT named | Gemini named | Named by both | ChatGPT’s also named by Gemini |
|---|---|---|---|---|
| “Best [service] in [town]” | 191 | 297 | 100 | 52% |
| Who can come today or this week | 258 | 286 | 112 | 43% |
| How much does it cost | 179 | 403 | 75 | 42% |
| “[Service] near [town]” | 234 | 321 | 96 | 41% |
| Cheapest with good reviews | 171 | 347 | 95 | 56% |
| One specific job (a hot tub, a shed) | 153 | 355 | 66 | 43% |
How many businesses each answer named
| Question | ChatGPT: businesses per answer | Gemini: businesses per answer | ChatGPT answers naming no business |
|---|---|---|---|
| “Best [service] in [town]” | 5.7 | 9.2 | 0% |
| Who can come today or this week | 6 | 9.5 | 5% |
| How much does it cost | 3.9 | 11.3 | 8% |
| “[Service] near [town]” | 5.6 | 10.2 | 5% |
| Cheapest with good reviews | 4.4 | 9.9 | 13% |
| One specific job (a hot tub, a shed) | 3.5 | 10.7 | 19% |
| All questions | 4.9 | 10.1 | 8% |
Same question, asked 3 times
| Engine | Named in all 3 answers | Named in 2 of 3 | Named in only 1 of 3 | Average overlap between two answers |
|---|---|---|---|---|
| ChatGPT | 16% | 22% | 61% | 31% |
| Gemini | 36% | 20% | 45% | 53% |
Where the answers got their information
| Source | ChatGPT: share of citations | Gemini: share of citations | ChatGPT: answers citing it | Gemini: answers citing it |
|---|---|---|---|---|
| Business websites (local companies and national chains) | 72.3% | 79.8% | 92% | 100% |
| Yelp | 0% | 0% | 0% | 0% |
| Other directories, lead sites and 'best of' lists | 13.6% | 8.7% | 39% | 63% |
| News, blogs and cost guides | 1.1% | 1.9% | 6% | 21% |
| Forums, social and video | 0.2% | 0.3% | 1% | 5% |
| Government sites | 9.4% | 4.6% | 35% | 53% |
| Other | 3.5% | 4.6% | 18% | 42% |
Which businesses got named
| Factor | Group | Businesses | Named by ChatGPT | Named by Gemini |
|---|---|---|---|---|
| Google reviews | Under 50 | 381 | 46% | 78% |
| Google reviews | 50–199 | 275 | 46% | 85% |
| Google reviews | 200+ | 247 | 62% | 86% |
| Google rating | Under 4.5 | 187 | 56% | 80% |
| Google rating | 4.5–4.7 | 62 | 56% | 79% |
| Google rating | 4.8+ | 654 | 48% | 83% |
| Google listing | Storefront address | 582 | 52% | 82% |
| Google listing | Service area only | 355 | 48% | 83% |
| Business type | Local business | 1,276 | 45% | 75% |
| Business type | Chain | 334 | 72% | 66% |
| Google listing found near the town | Yes | 937 | 50% | 82% |
| Google listing found near the town | No | 656 | 52% | 60% |
21 client websites: what the site has vs how often it was named
| Site has | Group | Sites | ChatGPT answers naming it (median) | Gemini answers naming it (median) |
|---|---|---|---|---|
| Pages in sitemap | Under 100 | 14 | 47% | 69.5% |
| Pages in sitemap | 100+ | 7 | 17% | 28% |
| Local guide pages | None | 7 | 39% | 67% |
| Local guide pages | 1+ | 14 | 39% | 55.5% |
| Town or service-in-town pages | Under 20 | 12 | 39% | 61% |
| Town or service-in-town pages | 20+ | 9 | 39% | 83% |
| llms.txt file | No | 9 | 50% | 67% |
| llms.txt file | Yes | 12 | 33.5% | 44.5% |
| Bing AI citations, Aug 11–Nov 10 | Under 50 | 11 | 39% | 61% |
| Bing AI citations, Aug 11–Nov 10 | 50+ | 9 | 39% | 83% |
What Gemini searched for behind the answers
| What the side search was about | Share of Gemini’s searches |
|---|---|
| Checking a named business | 63% |
| Reviews and 'best' lists | 32% |
| Price and cost | 26% |
| Specific items or jobs | 12% |
| Permits, rules and disposal | 11% |
| Availability and speed | 5% |
| Comparisons | 0% |
What the numbers say
Short version: being recommended by one doesn’t mean you’re recommended by the other. Fewer than half of ChatGPT’s picks showed up in Gemini’s answers to the same question, and 27% of Gemini’s showed up in ChatGPT’s. If you’ve only checked one of them, you’ve checked half the picture at best.
And don’t judge it from one search. ChatGPT gave a noticeably different list each time we asked. Only 16% of the businesses it named made all 3 answers. One screenshot from your office tells you almost nothing. Ask the same question a few times, from a private window, and from a town you serve that isn’t where your shop is.
What did line up: when a business got named, its own website usually got cited with it. That’s the part you control. Gemini also ran about 9 Google searches per answer on average, a lot of them checking a specific business by name or looking for prices. If your site doesn’t say what you charge, what you take and where you go, it has nothing to give those searches.
The other split: of the businesses either engine named, ChatGPT named 72% of the chains (same name in 3+ of our markets) and 45% of the local ones. Gemini named 75% of the locals. If you’re a local shop up against a franchise, Gemini was the friendlier of the two in this test. Reviews lined up too: ChatGPT named 62% of businesses with 200+ Google reviews, against 46% of those under 50.
And Yelp? Some answers mentioned it, but neither engine linked to it once in 756 answers. That’s the API, not the ChatGPT app, so don’t delete your Yelp page over it. It does mean websites were the main thing these answers cited.
One thing I’d be careful with: the 21 client sites table is small, and these are sites that already have work done on them. 18 of 21 were named by ChatGPT at least once for their own town and 20 by Gemini. Bigger sites and sites with an llms.txt file were named less often here, not more (median 17% of ChatGPT answers for 100+ page sites vs 47% for smaller ones). Bigger sites here may just be in bigger, more crowded markets, so I wouldn’t read that as “pages hurt.” It’s 21 sites. A pattern to watch, not a cause.
What I’d do this week: write down the 10 questions customers ask on the phone and make sure each one has a plain answer on your site, starting with price. Keep your Google Business Profile complete and your reviews coming. Then check both ChatGPT and Google, a few times each. The step-by-step version is in how to get your junk removal business recommended by ChatGPT.
Which pages AI tools actually cite is on AI citations by page type, and how Bing clicks line up with AI citations is on Bing clicks vs AI citations. Google’s Nov 12 switch to Gemini 4 Hydrogen is in the news post. For trade-specific playbooks see junk removal marketing and dumpster rental marketing. If you want us to do the work, that’s AI search SEO.
How we got this
- Questions: 126 local questions, 6 per market in 21 U.S. markets: “best [service] in [town]”, who can come today or this week, how much it costs, “[service] near [town]”, the cheapest with good reviews, and one question about a specific job each market really does. The markets are the main towns of 21 client businesses with live websites (mostly junk removal, plus demolition, HVAC and home organizing), so this isn’t a random sample of U.S. towns. The question set is fixed, so later runs can be compared like with like.
- Engines, both via API (not the ChatGPT or Google apps): OpenAI’s Responses API with web search (gpt-5-mini-2025-08-07) and Google’s Gemini API with Google Search grounding on Gemini 4 Hydrogen, the model Google put into Search on Nov 12, 2026. Each question ran 3 times on each engine on Nov 13, 2026. No location, account or history was attached, so a real person’s answers can differ.
- Named businesses: pulled from each answer by a separate model call that lists every business named, leaving out directories, review sites, government offices, landfills and charities. Name variants of the same business in a market were merged (same website, or one name the start of the other).
- Listings: each named business’s Google Maps listing (rating, review count, street address shown or service area only, website) looked up through Google’s Maps grounding in the Gemini API, 8 businesses per call. 2,106 businesses were looked up.
- Sources: every link an answer cited, grouped by type of site by a separate model call. Gemini reports its sources by domain only.
- Client sites: 21 client websites, one per market, crawled on Nov 13, 2026 (sitemap pages, local guide pages, town pages, llms.txt, homepage schema), joined with each site’s Bing AI citations (Bing Webmaster Tools, Aug 11–Nov 10, 2026).
- Not measured: Yelp ratings, review counts and whether a business has a Yelp page (Yelp blocks automated reading, and the search tools missed Yelp pages that exist), and Bing Places listings (no way to check them at scale). The one Yelp signal we could measure is whether answers linked to it as a source. Neither engine did.
- Caveats: one day of answers in 21 markets. The two sides aren’t the same size of model: OpenAI’s side is gpt-5-mini, the small model OpenAI’s API offers with web search, and ChatGPT itself may run a larger one; Google’s side is its new Gemini 4 Hydrogen. Correlation isn’t cause: a table showing named businesses had more reviews doesn’t show the reviews caused it. Anonymized: no business, town, region or website is named, here or in the CSV.
FAQ
Does ChatGPT recommend the same businesses as Google?
Mostly not. In 126 local home service questions across 21 markets (Nov 2026, through OpenAI’s and Google’s APIs), 46% of the businesses ChatGPT named were also named by Google’s Gemini for the same question.
Does ChatGPT give the same answer every time?
Not through the API (gpt-5-mini with web search). Asked the same question 3 times, it named 16% of its businesses in all 3 answers and 61% in only one. Gemini was steadier: 36% in all 3.
Where do ChatGPT and Gemini get local business recommendations from?
Business websites were the biggest source in both: 72.3% of ChatGPT’s citations and 79.8% of Gemini’s. Yelp was 0% of ChatGPT’s citations and 0% of Gemini’s.
Does ChatGPT favor chains over local businesses?
In this study, through the API, it leaned that way. Of the businesses either engine named, ChatGPT named 72% of the chains (same name in 3+ of our markets) and 45% of the local ones. Google’s Gemini went the other way: 75% of locals vs 66% of chains.
How do I check if ChatGPT recommends my business?
Ask it the questions your customers ask, a few times each, from a private window, and from a town you serve away from your shop. Do the same in Google. One search from your office tells you very little.
Related questions
Related reading
From the Desk of Justin Hubbard
One move a week, every Sunday.
Short, useful, written from inside a $2M home service company. Read by 2,000+ service business owners.