Which CRM do AI chatbots recommend? We asked 4 AI models 10 questions
More people now ask an AI assistant before they search Google. "What's the best CRM for a small business?" is exactly the kind of question that gets typed into ChatGPT, Claude, Gemini or Perplexity. So which CRMs come ba
More people now ask an AI assistant before they search Google. "What's the best CRM for a small business?" is exactly the kind of question that gets typed into ChatGPT, Claude, Gemini or Perplexity. So which CRMs come back?
We ran a small, repeatable test to find out. These are the results, including the parts that surprised us.
What we did
- 10 questions a buyer might actually ask, such as "What is the best free CRM?", "What CRM do you recommend for real estate agents?" and "Which CRM works best with Gmail and Google Workspace?"
-
4 models through their APIs: OpenAI
gpt-6-luna, Anthropicclaude-haiku-5.5, Googlegemini-3.5-flash-liteand Perplexitysonar(the only one that searches the web) - Each question asked twice per model, giving 80 answers in total
- 16 CRMs tracked: HubSpot, Salesforce, Zoho CRM, Pipedrive, Freshsales, monday CRM, Copper, Bigin, Attio, Less Annoying CRM, Insightly, Microsoft Dynamics 365, Capsule, Close, Keap and Nimble
We used default settings: no system prompt, no location and no web search, except that Perplexity always searches. For each answer we recorded whether each brand was mentioned, where it appeared in any list of recommendations, and which other names showed up. The test ran on November 16, 2026.
Finding 1: HubSpot shows up almost everywhere
| CRM | Mentioned in (of 80 answers) | Questions where it appeared (of 10) | Avg. list position when listed |
|---|---|---|---|
| HubSpot | 99% | 10 | 1.9 |
| Pipedrive | 64% | 9 | 2.6 |
| Salesforce | 57% | 10 | 3.9 |
| Zoho CRM | 51% | 9 | 2.5 |
| Freshsales | 28% | 7 | 3.5 |
| monday CRM | 16% | 6 | 3.8 |
| Copper | 14% | 3 | 2.3 |
| Bigin | 10% | 5 | 2.5 |
| Attio | 9% | 2 | 2.8 |
| Less Annoying CRM | 9% | 4 | 3.4 |
| Insightly, Dynamics 365, Capsule | 5% each | 2-4 | - |
| Close | 2% (see note) | 1 | - |
| Keap, Nimble | 0% | 0 | - |
HubSpot appeared in 79 of 80 answers. It was the first of our tracked CRMs to be named in 72% of answers, and it was the #1 item in the answer's list 36 times. The only answer without it was one Perplexity answer to the real-estate question.
After the top four (HubSpot, Pipedrive, Salesforce, Zoho), visibility drops off sharply. Freshsales appeared in about one answer in four, and everything else in fewer than one in six.
Note on Close: we tracked it as "Close CRM" so we wouldn't count the everyday word "close." Answers often list it simply as "Close", and counting those, it appeared in 11 of 80 answers (14%).
Finding 2: The four models disagree more than you'd expect
Share of each model's 20 answers that mentioned the brand:
| CRM | OpenAI gpt-6-luna | Claude Haiku 5.5 | Gemini 3.5 Flash Lite | Perplexity Sonar |
|---|---|---|---|---|
| HubSpot | 100% | 100% | 100% | 95% |
| Pipedrive | 60% | 70% | 55% | 70% |
| Salesforce | 50% | 70% | 70% | 40% |
| Zoho CRM | 45% | 75% | 15% | 70% |
| Freshsales | 15% | 20% | 20% | 55% |
| monday CRM | 5% | 10% | 30% | 20% |
| Bigin | 5% | 0% | 5% | 30% |
Zoho CRM is the clearest example. Claude and Perplexity named it in about three of four answers, but Gemini named it in only 3 of 20. Freshsales did much better on Perplexity, the model that searches the web, than on the three that answer from memory.
If you only check one assistant, you can get a very different picture of where your brand stands.
Finding 3: The question decides the winner
The averages hide big swings between questions:
- "Which CRM works best with Gmail and Google Workspace?" Copper was the first tracked CRM named in 7 of 8 answers, compared with 14% visibility overall. Streak, a Gmail-based CRM we didn't track, appeared in 6 of the 8 answers.
- "What CRM do you recommend for real estate agents?" The lists were led by real-estate tools we didn't track, including Wise Agent (7 of 8 answers), Lofty (6) and Follow Up Boss (at least 4). HubSpot still appeared, but usually as a "general-purpose option" around position 4.
- "What is a good affordable alternative to Salesforce?" Salesforce was named first in 5 of 8 answers because the question names it. HubSpot was #1 in the list in 6 of 8, and Zoho CRM was mentioned in all 8.
- "What is the best free CRM?" It came down to two brands: HubSpot (8 of 8) and Zoho CRM (7 of 8).
For a niche product, broad questions ("best CRM for small business") are mostly a contest between the big names. Specific questions are where smaller brands win, but only if AI already associates them with that niche.
What this means if you sell software
- Check more than one model. Visibility for the same brand ranged from 15% to 75% across models.
- Test the questions your buyers ask, not your category name. Copper is nearly invisible on general CRM questions and the default answer on the Gmail question.
- Look at who's in the list besides your known competitors. Streak, Wise Agent, Lofty and Close came up often without being on our list. AI assistants can surface competitors you don't track.
- Re-check over time. One snapshot is a starting point, not a verdict (see the caveats below).
How to check your own brand
We ran this study with our AI Brand Visibility Tracker on Apify, and you can run the same check for your brand. Enter your brand, a few competitors and the questions your customers ask. For each answer it returns whether you were mentioned, your list position, which competitors appeared and the sentence the AI used about you, plus a visibility score and share of voice per model.
You don't need an AI API key or a subscription. It costs $0.08 per AI check (one question answered by one model), so a 10-question, 4-model check like the one above, asked once, costs $3.20.
Method notes and caveats
- AI answers vary from run to run. We asked each question twice per model, but 80 answers is still a small sample. Treat differences of a few percentage points as noise.
- These are API models, not the consumer apps. We used cost-efficient API models. ChatGPT, Gemini and other apps may use larger models, personalize answers or search the web, so what a logged-in user sees can differ.
- Detection is name-based. A brand counts as "mentioned" if its name or alias appears in the answer. We picked aliases to avoid false matches (see the note on Close), which can undercount brands with generic names.
- "List position" only applies when the answer contains a list. 75 of the 80 answers did.
- Some answers were cut off. Answers were capped at about 1,000 tokens, and 17 of 80 hit that limit. Brands that would have appeared near the end of those answers are undercounted.
- One question names Salesforce. That inflates Salesforce's "named first" numbers for that question.
- This is a snapshot from November 16, 2026. Models are updated often, and results can change.
The raw results (one row per answer per brand, with snippets) are available as a CSV.
Signal Wharf builds small data tools for marketers and researchers.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.