We asked ChatGPT and Perplexity which sales copilot to buy
We asked ChatGPT and Perplexity what to buy. Who they named, why page count beat quality, and how to shortlist a tool.
AI doesn’t pick the best product. It picks the best documented one. That is our read of 25 answers, not a law of nature, and we will show the counts so you can disagree.
We make Axelize. We ran this because buyers now ask ChatGPT and Perplexity which tool to buy, and we wanted to know why they rarely named us. So this post is a vendor reading its own bad grades out loud. Take the numbers with that in mind, and with the limits below.
What we ran, and what is wrong with it
- ChatGPT: three accounts, each with five buyer prompts typed one after another in a single chat. Fifteen answers. Memory was on in all three.
- Perplexity: one thread with ten buyer prompts, one after another. Ten answers.
- The prompts were things a seller types: “best AI sales assistant that doesn’t join the meeting as a bot”, “cheapest real-time AI sales assistant”, “Convo AI alternatives”, “I freeze on sales calls when a prospect pushes back on price. Is there an AI that can help me live?”
The limits, bluntly. It is a small sample. Typing prompts in sequence lets earlier answers shape later ones. Memory was on, which we show below is not a small thing. Treat every count as a direction, not a rate.
Who they named
Counts are the number of answers that name the tool at all, from our own reading of the answers.
| Tool | ChatGPT, of 15 | Perplexity, of 10 |
|---|---|---|
| SalesGhost | 12 | 7 |
| CallCopilot | 8 | 1 |
| Convo | 2 | 8 |
| Axelize | 6 | 0 |
| NextSay | 4 | 0 |
| Attio | 2 | 4 |
| HubSpot | 2 | 4 |
| Clari Copilot | 0 | 5 |
| Gong | 3 | 2 |
| Othello | 3 | 1 |
Convo’s two ChatGPT mentions are both answers to the prompt that named Convo. Without that prompt it would be zero. Perplexity named Convo first in six of its ten answers.
0 of 10
Six things we saw
1. The engines disagree. SalesGhost is the only tool near the top of both lists. ChatGPT leaned on SalesGhost, CallCopilot and NextSay. Perplexity leaned on Convo, Attio and Clari Copilot. If you ask one engine you get one world.
2. The pool is mostly pages the vendors wrote. In ChatGPT’s answers, SalesGhost’s own site held 36 of 187 source slots. In Perplexity’s, Convo’s own site held 42 of 367. Both publish a page for a lot of the questions a buyer might type: alternatives, pricing, a bot-free version, a Google Meet version. Nothing we saw suggested the engines checked whether a product works. They matched the question to the page.
3. Free plans win price questions. In the “cheapest” answers, the tools named first had a free plan. Axelize, which has a trial and no free plan, came fourth or last every time it was named.
4. “No recording” sends the model to privacy pages. Ten of ChatGPT’s 25 searches used recording or audio words, and it pulled vendors’ security and privacy pages to check. A tool with no such page loses that round by default.
5. They quote the cheapest number. Both engines quoted Convo at $22.49 a month, and most answers did not say which plan that is. We read its pricing page: it is the Starter plan, which lists “real-time insights”, while the Live Assistant, the real-time answers, sits on a $52.49 plan billed annually. One ChatGPT answer did add that full live assistance may need a higher tier. The others did not.
6. Memory leaks into the answer. In one chat, ChatGPT wrote “Axelize” and “captions only” into its own search query on the second prompt, before anything in that chat had mentioned either. The likely cause is account memory. It means your results on your own account are not what a buyer sees, and it means our ChatGPT counts above are probably too high.
How to shortlist with AI without getting played
The engines are good at one thing: collecting candidates fast. They are bad at telling you which plan has the feature, or which page is old. So split the job. Let the engine build the list, then check the list yourself.
A better shortlist prompt
I sell on [Google Meet, Zoom or Teams]. I need help during calls, not notes afterwards. My limits: [no bot in the meeting], [no audio stored], [budget per user per month], [CRM it must sync with].
List up to six tools. For each one, give:
- The cheapest plan that includes live suggestions during the call, and its price
- Whether a bot joins, and what is stored (audio, transcript)
- The page you got each fact from, and the date on that page
Say “not stated” where a page does not say. Do not use my past chats.
Then run it in a second engine and keep the names that appear in both. Then open each vendor’s pricing and security page yourself, because those are the pages the engines were reading in the first place.
Check any AI answer before you buy
- Run the question in a temporary chat with memory off
- Ask for the plan that includes the feature, not the cheapest card
- Open the vendor’s pricing and security pages and read the date
- Run it in a second engine and keep only the names in both
- Treat any “best of” list written by a vendor, including this one, as marketing with sources
What this means if you are a buyer
Use the engines for what they are good at. Then try two tools on real calls. A trial costs less than the hour you spend arguing about which list to trust. If you want the live sentence written in your voice and you sell on Google Meet, Axelize has 14 days or 5 calls free with no card, then plans from $40 a month. It works with Google Meet. Zoom and Microsoft Teams are on the roadmap. If you are shopping for a free plan, SalesGhost and CallCopilot have one, and we said so above.
If you are still choosing a sales tool from one chatbot answer in 2026, that is not research, that is a vibe check. A twenty-minute trial beats it.