Who AI recommends: the firms ChatGPT, Gemini and Google AI name in 22 B2B categories (October 2026)
We asked five AI engines 264 buyer questions across 22 B2B categories, each phrased the way a buyer asks for a provider, and counted the firms every answer named. 1,232 of 1,249 answers (98.6%) named at least one firm. In every category the most named firm was named by all five engines and appeared in 21.8% (Solar Landscape, commercial solar installers and EPC contractors) to 93.0% (The Hartford, small business insurance providers) of answers, while most other firms were named in only one answer.
By Frederik Jakobsen, founder of Danish Lead Co and SourceRank AI. Measured 10 and 11 October 2026, published October 2026.
Who AI named, category by category
Each category is its own page with the full firm table, engine by engine results, cited sources and every question asked. Answers excludes Copilot answers written for a buyer outside the US (a side effect of how we access Copilot) and Google searches with no AI Overview. The first ten categories were measured on 10 October 2026 and the other twelve on 11 October 2026.
PE, M&A and finance
| Category | Answers | Named a provider | Distinct firms | Most named firm | Top 3 share of mentions | Measured |
|---|---|---|---|---|---|---|
| Lower middle market private equity firms | 57 | 54 of 57 94.7% | 200 | The Riverside Company 19 of 57 (33.3%) | 11.3% | 11 October |
| Sell-side M&A advisors (lower middle market) | 58 | 52 of 58 89.7% | 140 | Houlihan Lokey 26 of 58 (44.8%) | 15.4% | 11 October |
| Middle-market investment banks | 59 | 58 of 59 98.3% | 148 | Houlihan Lokey 50 of 59 (84.7%) | 22.2% | 11 October |
| Quality of earnings and financial due diligence providers | 60 | 60 of 60 100.0% | 141 | BDO USA 34 of 60 (56.7%) | 14.5% | 10 October |
| Fractional CFO and outsourced accounting firms | 59 | 59 of 59 100.0% | 114 | Kruze Consulting 34 of 59 (57.6%) | 24.1% | 10 October |
| Revenue-based financing and lenders for SaaS companies | 55 | 54 of 55 98.2% | 81 | Founderpath 44 of 55 (80.0%) | 27.3% | 11 October |
Insurance, benefits and HR
| Category | Answers | Named a provider | Distinct firms | Most named firm | Top 3 share of mentions | Measured |
|---|---|---|---|---|---|---|
| Employee benefits brokers and consultants | 58 | 58 of 58 100.0% | 74 | Marsh McLennan Agency 37 of 58 (63.8%) | 25.5% | 11 October |
| Commercial insurance brokers | 55 | 55 of 55 100.0% | 89 | Marsh McLennan 38 of 55 (69.1%) | 26.8% | 11 October |
| Small business insurance providers | 57 | 57 of 57 100.0% | 33 | The Hartford 53 of 57 (93.0%) | 46.1% | 11 October |
| PEO providers | 58 | 58 of 58 100.0% | 37 | ADP TotalSource 50 of 58 (86.2%) | 45.2% | 11 October |
Industrial, energy and logistics
| Category | Answers | Named a provider | Distinct firms | Most named firm | Top 3 share of mentions | Measured |
|---|---|---|---|---|---|---|
| CNC machining and contract manufacturers | 54 | 54 of 54 100.0% | 175 | Xometry 32 of 54 (59.3%) | 26.0% | 10 October |
| Industrial automation system integrators | 56 | 55 of 56 98.2% | 210 | JR Automation 16 of 56 (28.6%) | 9.8% | 11 October |
| Commercial solar installers and EPC contractors | 55 | 53 of 55 96.4% | 169 | Solar Landscape 12 of 55 (21.8%) | 8.4% | 11 October |
| Commercial roofing contractors | 50 | 49 of 50 98.0% | 115 | Tecta America 36 of 50 (72.0%) | 32.8% | 10 October |
| 3PL and ecommerce fulfillment companies | 58 | 58 of 58 100.0% | 68 | ShipBob 52 of 58 (89.7%) | 40.8% | 10 October |
| Freight brokers | 58 | 58 of 58 100.0% | 98 | C.H. Robinson 47 of 58 (81.0%) | 29.4% | 11 October |
Technology and services
| Category | Answers | Named a provider | Distinct firms | Most named firm | Top 3 share of mentions | Measured |
|---|---|---|---|---|---|---|
| Managed IT service providers | 56 | 54 of 56 96.4% | 152 | Dataprise 33 of 56 (58.9%) | 21.1% | 10 October |
| Managed detection and response (MDR) providers | 56 | 56 of 56 100.0% | 42 | Arctic Wolf 46 of 56 (82.1%) | 41.3% | 10 October |
| Field service management software | 56 | 56 of 56 100.0% | 54 | ServiceTitan 44 of 56 (78.6%) | 32.9% | 10 October |
| B2B PPC and paid media agencies for SaaS | 59 | 59 of 59 100.0% | 85 | Directive Consulting 44 of 59 (74.6%) | 27.1% | 10 October |
| Executive search firms for technology and B2B companies | 57 | 57 of 57 100.0% | 96 | Spencer Stuart 35 of 57 (61.4%) | 24.8% | 11 October |
Healthcare
| Category | Answers | Named a provider | Distinct firms | Most named firm | Top 3 share of mentions | Measured |
|---|---|---|---|---|---|---|
| Medical billing and revenue cycle management companies | 58 | 58 of 58 100.0% | 142 | athenahealth 33 of 58 (56.9%) | 18.5% | 10 October |
Top 3 share of mentions: of every time a firm was named in that category, the share that went to the three most named firms. Measured: the date of the single run behind that category (2026). Summed over the 22 categories, 2,463 distinct firms were named; a firm named in two categories counts twice.
What held across all 22 categories
- +AI engines almost always name firms. 1,232 of 1,249 answers (98.6%) named at least one provider, an average of 7.0 firms per answer. A buyer who asks gets a shortlist, not a list of criteria. The least willing category was sell-side M&A advisory, where 52 of 58 answers named an advisor.
- +Every category has a leader that all five engines name, but how far ahead it is varies widely. The most named firm's share of answers ran from 21.8% (Solar Landscape, commercial solar installers and EPC contractors) to 93.0% (The Hartford, small business insurance providers).
- +Below the leaders the lists fragment. Of the firms named in each category, 1,479 were named in only one answer, summed across the 22 categories. In industrial automation 135 of 210 firms appeared once.
- +Concentration differs by a factor of about 5. The three most named firms took 46.1% of all firm mentions for small business insurance providers and 8.4% for commercial solar installers and EPC contractors. The most fragmented were commercial solar installers and EPC contractors, industrial automation system integrators and lower middle market private equity firms; the most concentrated small business insurance providers, PEO providers and managed detection and response providers.
- +Review directories are not where most of this comes from. Of the 814 answers that showed sources, 111 (13.6%) cited a directory, review or analyst site, concentrated in managed IT (Clutch) and MDR and field service software (Gartner); in the twelve categories added on 11 October it was 29 of 416. A named firm's own website was cited in 609 (74.8%).
How willing each engine is to name names
Gemini named the most firms per answer (8.2 on average), Google AI Mode the fewest (5.4). All five engines named at least one firm in nearly every answer.
| Engine | Answers | Named a provider | Firms per answer, average | Answers with a source list |
|---|---|---|---|---|
| ChatGPT | 264 | 262 of 264 99.2% | 7.9 | 0 of 264 |
| Gemini | 264 | 263 of 264 99.6% | 8.2 | 204 of 264 |
| Microsoft Copilot | 200 | 196 of 200 98.0% | 7.4 | 200 of 200 |
| Google AI Mode | 264 | 261 of 264 98.9% | 5.4 | 158 of 264 |
| Google AI Overviews | 257 | 250 of 257 97.3% | 6.1 | 252 of 257 |
All 22 categories pooled. Firms per answer is the average number of distinct provider firms named per measured answer, including answers that named none.
Exactly what was measured
| Parameter | Value |
|---|---|
| Date | 10 October 2026 for the first ten categories and 11 October 2026 for the twelve added in the second batch. One run per category |
| Categories | 22 B2B categories, measured in two batches with the same engines, question format and method. We left out our own category (AI search, SEO and outbound agencies). |
| Buyer questions | 264, 12 per category, each phrased as a buyer asking for a provider. Every question is listed on its category page |
| Answer engines | 5: ChatGPT, Gemini, Microsoft Copilot, Google AI Mode, Google AI Overviews |
| Engine cells | 1,320. One question on one engine |
| Answers measured | 1,249 of 1,320. 7 Google searches showed no AI Overview, and 64 Microsoft Copilot answers came back written for a buyer outside the United States (for example in Riyadh, Pune, Rosario or São Paulo), a side effect of the third-party service we query Copilot through, which runs each session from a random location; both are excluded. First batch: 566 of 600; second batch: 683 of 720. Every engine call that failed was retried, and none was left unanswered |
| Market | United States, English. Google AI Mode and AI Overviews were queried with a US location. ChatGPT, Gemini and Copilot were not given a location, the same as in our client measurement runs |
| Quality checks | Every answer was checked against the question it was filed under. 18 answers that answered a different question, merged two questions, returned an error or refusal, or returned only a place name were set aside and asked again one at a time: 10 in the first batch (9 from ChatGPT and 1 from Gemini) and 8 in the second (5 from Gemini and 3 from ChatGPT), where ChatGPT was queried at a lower concurrency. In the second batch, a firm that appeared in a ChatGPT answer only as a cited source label was not counted |
| Sources | Cited sources come from Gemini, Copilot, Google AI Mode and Google AI Overviews. ChatGPT's answers in both runs came back without a machine-readable source list |
| Firm extraction | One language model call per answer listed the provider firms it named, excluding directories, publications, citation labels and software tools. Names were normalised and aliases merged by rule and by hand. Two hand checks of 20 randomly drawn answers each, one per batch, compared the firm list with the raw text. First batch: 19 of 20 exactly right, 125 firms extracted with none wrong and 2 missed. Second batch: 17 of 20 exactly right, 150 firms extracted, 1 counted that the answer named only as unsuitable, and 3 missed |
| Unit of the share | An answer. A firm counts once per answer however often the answer repeats it |
| Not measured | Location personalisation for the chat engines, logged-in or personalised sessions, run-to-run variation (one run per category; the same question asked again can return different firms), and model versions, which change without notice. The two batches were measured a day apart, so a difference between a first-batch and a second-batch category can include a day of engine drift |
The engines are queried the same way as in our client measurement runs, described in the benchmark methodology. We did not contact or pay any firm named here, and no firm saw these results before publication.
Cite this research
Journalists, analysts and writers are welcome to quote, chart and republish the figures on this page. Please credit SourceRank AI and link to this page as the source. For the underlying data, a comment from Frederik Jakobsen or a category we have not covered, see the press page.
Suggested citation
SourceRank AI (2026). Who AI recommends: the firms ChatGPT, Gemini and Google AI name in 22 B2B categories. Measured on 10 October 2026 (ten categories) and 11 October 2026 (twelve categories). https://sourcerankai.com/research/who-ai-recommends/
Link to copy
<a href="https://sourcerankai.com/research/who-ai-recommends/">Who AI recommends: the firms ChatGPT, Gemini and Google AI name in 22 B2B categories (SourceRank AI, October 2026)</a>If your firm is not on the list
The pattern across categories is consistent. Leaders are named by every engine whether or not their own pages are cited. Challengers that got named did it by publishing specific comparison and situation pages that the search-grounded engines (Copilot, Google AI Mode and AI Overviews) read and cite: SHIPHYPE in 3PL, SaaSHero in paid media, Founderpath in SaaS lending, Talentfoot in executive search, Windsor Drake in sell-side M&A. ChatGPT returned no source list in either run, so which pages it relied on cannot be seen from the outside. That is the work described on the services page. The free site readiness check looks at whether your site is structurally ready to be cited; it does not measure answers. To see where your firm stands in your own category, book a call.
Where to go next
What AI engines actually cite
One full measurement run taken apart domain by domain: which sources the five engines read.
What AI engines cite by industry
The source mix behind AI answers in ten industries, from our September 2026 runs.
The measurement method
How tracked question sets, engines and snapshots work.
Common questions
How often does AI name a specific firm when a buyer asks for a provider?
Almost always. In this October 2026 study 1,232 of 1,249 answers (98.6%) from ChatGPT, Gemini, Microsoft Copilot, Google AI Mode and Google AI Overviews named at least one provider firm, across 22 B2B categories.
Do different AI engines recommend the same firms?
At the top, yes: in all 22 categories the most named firm was named by all five engines. Below the top few names they diverge, and 1,479 firm names, summed across the categories, appeared in only one answer.
Do review sites like Clutch and G2 decide who AI recommends?
Less than most people assume. Of the 814 answers that showed a source list, 111 (13.6%) cited a directory, review or analyst site, mostly Clutch for managed IT and Gartner for MDR and field service software. A named firm's own website was cited in 609.
How can a firm get recommended by ChatGPT and Google AI?
By being written about specifically and consistently where the engines read: the firm's own pages answering the buyer's exact situation, third-party roundups and comparisons in the category, and, in some categories, review and analyst sites. SourceRank AI, the answer engine optimization service from Danish Lead Co, measures this per category and does the work.