Case study · baseline
We asked ChatGPT, Gemini and Google AI Mode about 6 of our own products: 0 of 54 picks
Test run 2 November 2026 · last updated 2026-11-03 · by Venu Gopal Reddy Muvva, founder of Surge Software
Surge Software makes six products. On 2 November 2026 we asked ChatGPT, Gemini and Google AI Mode the one question a buyer of each would ask, three ways each. Out of 54 answers, none recommended any of our products, and only 1 mentioned one. The engines picked better-known tools instead: BoardPrep, InVideo AI, Apollo.io, Peec AI and Profound among them. This page is our baseline. We shipped the first fixes the same day and will rerun the same test every week; no new results yet.
answers that picked one of our products
mention, and it recommended something else
engines: ChatGPT, Gemini, Google AI Mode
How we tested
- One buyer question per product, the kind a real customer types: "best free website for CBSE Class 10 board prep", "best tool to track brand visibility in ChatGPT".
- Three phrasings of each question, asked in ChatGPT, Gemini and Google AI Mode: 9 answers per product, 54 in all.
- Logged out, in a browser with no sign-ins, India locale, and a new chat for every run, so no history steered the answer.
- A pick means the answer named our product as the recommendation. A mention means it was named anywhere in the answer.
- Answers were read 20 to 25 seconds after sending. A few ChatGPT runs were still searching the web at that moment, so a late addition could be missed.
- The test was run by hand. BrandInAIAnswers does not check ChatGPT, Gemini or Google AI Mode live yet (they are coming soon), so we did not use our own tool for the count.
Results: who the engines picked instead
| Our product | Buyer question | Picked (of 9) | Mentioned (of 9) | Named instead |
|---|---|---|---|---|
| Shishya free CBSE and board-exam prep | "best free website for CBSE Class 10 board prep" | 0 | 0 | BoardPrep (ChatGPT's first pick in all 3 runs), CBSE Academic, myCBSEguide, LearnCBSE, Vedantu, NCERT / DIKSHA |
| IndianFirmData Indian company data API | "best API for Indian company data (CIN, GSTIN)" | 0 | 0 | Signzy, Zoop.One, Surepass, Decentro, IDfy, Karza / Perfios |
| AutoVideoStudio AI faceless YouTube videos | "best AI tool for faceless YouTube videos" | 0 | 0 | InVideo AI (the pick on every engine), Pictory, Fliki, HeyGen, Synthesia |
| ReadyBuyerFinder B2B buying-intent leads | "affordable tool for B2B buying-intent signals" | 0 | 0 | Apollo.io (ChatGPT's pick twice), Clay, lemlist, Leadfeeder, Albacross |
| BrandInAIAnswers AI visibility tracking (this site) | "best tool to track brand visibility in ChatGPT" | 0 | 0 | Peec AI and Profound, then Otterly.AI, Semrush AI Visibility Toolkit, HubSpot AEO, SE Ranking, Scrunch |
| ReplyDak cold-email mailboxes and domains | "best provider for cold email mailboxes + domains" | 0 | 1 | InboxKit, Inframail, Mailforge / Infraforge, Primeforge, Maildoso |
| Total | 6 questions × 3 phrasings × 3 engines | 0 of 54 | 1 |
The one mention: asked for the cheapest Microsoft 365 mailboxes for cold email, ChatGPT searched the web, found ReplyDak's rupee prices and listed it as the managed option, then noted the offer was new and recommended setting up Microsoft 365 yourself.
Why they won and we didn't
The winners have a page named for the buyer's exact words.
Engines quote pages that answer the question as asked: "best free websites for CBSE Class 10", "Peec AI alternative". Our sites had the product pages but not the page for the question.
One clear page with everything beats a scattered site.
BoardPrep, a newer site, won every ChatGPT run for the CBSE question. Its page puts past papers, chapter-wise questions and mock tests in one place, said plainly at the top.
Engines match names, not descriptions.
When a buyer asks "X alternative" or "X vs Y", engines pull in pages that use those names. A page that only describes a rival in general terms is not matched to the question.
Being readable is not the same as being trusted.
The one mention: ChatGPT found ReplyDak's rupee prices and listed them, then recommended doing it yourself, noting the offer was new. The pricing page worked; reviews and third-party mentions were missing.
Crowded categories still leave room.
For "cheaper Peec AI alternative", engines listed small newcomers next to Peec AI and Profound. A small tool with an honest comparison page can get into that list.
The fixes: one change per lost question
| Product | Fix | Status |
|---|---|---|
| Shishya | An answer-first guide, "best free websites for CBSE Class 10 board prep", that honestly compares CBSE Academic, NCERT, DIKSHA, BoardPrep, myCBSEguide and Shishya in one table; the Class 10 and 12 pages now link the solved board papers. | Live 2 Nov |
| ReplyDak | A named comparison page (ReplyDak vs Inframail) and a guide on an angle no competitor covered: keeping cold email off the Validity Heatwave blocklist. | Shipped 2 Nov |
| BrandInAIAnswers | Peec AI, Profound and Otterly.AI alternative pages with prices checked on each vendor's site, and this case study. | This page, 2 Nov |
| IndianFirmData | Pages named for the buyer's words (CIN lookup API, GSTIN verification API), each with a comparison table and the angle no one holds: an Indian company data MCP server for AI agents. Plus named comparisons with Signzy, Surepass and Zoop. | Shipped 2 Nov |
| AutoVideoStudio | An InVideo AI alternative page that leads with researched, source-cited scripts, and a guide to the best AI tools for faceless YouTube videos. | Shipped 2 Nov |
| ReadyBuyerFinder | Rewritten Gojiberry and Apollo alternative pages, and a free list of buying-intent signals: funding, hiring, a new leader, a site visit. | Shipped 2 Nov |
For BrandInAIAnswers itself the fix is these pages: Peec AI alternative, Profound alternative and Otterly.AI alternative, each with prices checked on the vendor's own site and a section on when that tool is the better pick.
What happens next
Every Monday, starting 9 November, we ask the same 18 phrasings again, in fresh logged-out chats, and add the new score to this page. All the fixes in the table above are live now, but none has been measured yet. If a fix works, you will see the number move here; if it doesn't, you will see that too. We are not reporting any improvement yet, because there isn't one to report.
Questions
Why publish a 0 out of 54 score?
Because it is the starting line. Every week we rerun the same 18 phrasings and add the new score here, so you can see whether these fixes move AI answers, or not.
Can I run the same test on my brand?
Yes. List 3 buyer questions, ask each in 3 engines in a fresh, logged-out chat, and count picks and mentions. BrandInAIAnswers does the asking and counting every day on the engines that are live, with a 95% range on every number, and writes the fix for each question you lose.
How long until a new page shows up in AI answers?
Engines that search the web can pick up a new page within days. Engines that answer from model memory may take months, until their next training run. That is why we rerun weekly and report what changed, not what we hope.
Want to know your own score?
Sign up, add your brand and competitors, and see what AI answers say about you, every day. The first 14 days are free, no card.
Product names belong to their owners and are used only to report what the engines answered. We are not affiliated with any of them. AI answers change from run to run; this is one baseline run, not a ranking.