One overall rank, many buyer questions
DecaGEO does not ask ChatGPT “what’s the best CRM?” Every question in its weekly panel describes a buyer, with one to three conditions such as company size, main use case or the integrations the buyer needs, and more than four in five combine two or three. A typical Marketing Automation question reads: “What are the best email-focused marketing automation tools for a small startup that just needs simple automated email workflows without a lot of complexity?” A brand’s overall rank comes from its DECA Score, which adds up the brand’s recommendations across all of those questions and weights each by the brand’s place in the answer. A brand that ChatGPT puts first for one kind of buyer and leaves out for the others collects points from a few questions and ends up low on the overall list. DECA Score sets each week’s category leader at 100, so other brands’ scores also move when the leader’s does. Whether a brand itself was recommended more often shows in its Mention Rate.How many brands outside the top 10 still come first for a buyer
DecaGEO, ChatGPT (GPT-5.6 Terra), US, English, 10 B2B software categories, four weekly runs. A brand ranked in two categories counts once in each. Definitions in How we measured.
Seven in ten of these brands reached the first three of some answer at least once. Holding that place for the same buyer in all four weeks is rarer, and 67 did. They were spread across nine of the ten categories, with the most in Project Management (17), SEO (15) and Email Marketing (13).
Three brands that rank low overall and high for one buyer
MailerLite, Marketing Automation, 11th at best. Five questions put MailerLite among the first three in all four weeks, and in each of them it was first in at least one week. All five describe a small startup with basic email automation needs, such as scheduled campaigns with A/B testing, onboarding and reactivation workflows, or simple automated sequences. TeamGantt, Project Management, 22nd to 30th. From August 9 to September 27, 2026, a small startup asking for a simple Gantt and timeline tool to track client deliverables, resource allocation and milestone dependencies got TeamGantt first in all eight weekly answers. The September 6 answer opened with “TeamGantt is probably the best fit if your startup’s priority is a simple, schedule-first tool”. Two more Gantt questions from small startups had it in the first three in every one of those weeks. Similarweb, SEO, 33rd to 50th. A mid-market company with an in-house SEO team asking for keyword research and competitive analysis platforms got Similarweb third in all eight weekly answers. The August 9 and September 27 answers listed it as “Best for market-level competitive intelligence”. TeamGantt and Similarweb are two of 16 brand-question pairs from a longer, eight-week window, counted per question and separate from the 16 brands in the table’s last row. Each brand was ranked below 20th in every week from August 9 to September 27, 2026, and sat in the first five of the same question’s answer in every one of those weeks. In each of the four answers from September 6 to 27, the passage around the brand’s first mention recommends it rather than naming it in passing.
For 16 brand-question pairs, the brand's overall category rank and its place in the answer to one buyer question, lowest to highest over eight weeks. DecaGEO, ChatGPT (GPT-5.6 Terra), US, English, weeks of August 9 to September 27, 2026.
The answers behind these wins named four or five products
The 60 weekly answers behind the 16 pairs named a median of four recommended products, and the 384 answers behind the 67 brands’ four-week wins named a median of five. Across every answer in the ten categories in the same four weeks, the median was also five. Answers naming three products or fewer made up 13% and 15% of the two sets, against 17% and 16% of all answers weighted to the same category mix.What a September 2026 study found with search turned off
A preprint posted to arXiv on September 14, 2026, by Edward Malthouse and co-authors tested six large language models on five consumer product categories, from cordless drills to cat food, with web search and other retrieval tools turned off. Plain category questions left out many established brands. When the question described the buyer’s goals and constraints, the set of recommended brands changed, and brands left out of ordinary recommendations could still come back when the prompt supplied distinctive cues. The authors conclude that “low recommendation rates do not necessarily reflect a lack of LLM knowledge about the brand.” The DecaGEO figures cover a different setting, ChatGPT recommending B2B software to buyers who describe themselves. In them, a low overall rank and a first-place answer for a specific buyer can belong to the same brand in the same week.How to check your own brand
- Find your overall place. The public category ranking shows every ranked brand’s rank, weekly movement and Mention Rate for the latest week, without signup. If your brand is not in your category’s ranking at all, the figures here and steps 2 to 5 do not apply to it yet, and Break into the ranking covers that case.
- Open the Segment Position tab. Once you sign in, including on the free plan, your product page shows a Segment Position card for each buyer condition in your category, such as each business scale, use case or integration need. Each card gives your rank among the answers to that condition’s questions and the share of those answers that recommend you. Conditions where ChatGPT did not recommend you show “Not mentioned”. Look for cards where your rank is well above your overall rank.
- Click the card and read the questions. The card opens Prompt & AI Response filtered to that condition’s answers for the week. Each row is one buyer question, and its Brand Ranking Trail lists the recommended brands in answer order with yours highlighted. The Trail shows only brands the answer recommends, while the place in the answer used in the figures above counts every product the answer names, so the two can differ. Note which questions put you near the top and what buyer each one describes.
- Check the same questions in earlier weeks. The week selector at the top of the product page switches every tab, including this one. The free plan covers the last four weeks, and Pro covers ten. A question that keeps you near the top week after week is a buyer ChatGPT consistently matches you to.
- Compare those buyers with the ones you sell to. If the questions you win describe your customers, your overall rank understates what those buyers see. The cards marked “Not mentioned” or with a low rank are the buyers whose answers leave you out, and Break into the ranking walks through checking whether your pages describe those conditions in the words ChatGPT uses. If no card puts you near the top in any week, start with that guide’s first step.
FAQ
If my brand isn’t in ChatGPT’s top 10 for my category, does ChatGPT not recommend it?
Not necessarily. From September 6 to 27, 2026, 67 of 509 brands ranked outside their category’s top 10 in DecaGEO’s ChatGPT rankings were among the first three brands in the answer to the same buyer question in all four weeks. See the table.Why can a brand rank low overall but first for one buyer?
The overall rank adds up recommendations across many buyer questions. A brand that ChatGPT puts first for one kind of buyer and leaves out for others scores on few questions. See One overall rank, many buyer questions.Did these brands rank high only because the answers were short?
The answers behind the brands’ wins named a median of four or five recommended products, and all answers in the same weeks named a median of five. See The answers behind these wins named four or five products.How do I find which buyers ChatGPT recommends my brand to?
Sign in, open your product’s Segment Position tab, click the cards where you rank highest, and check each question’s Brand Ranking Trail in Prompt & AI Response across the weeks you can select. See How to check your own brand.How we measured
The panel’s questions, schedule and metric definitions are on the methodology page. What is specific to these figures:- Window. Four weekly runs, weeks of September 6 to 27, 2026, all on GPT-5.6 Terra, ten categories, the same questions every week. The eight-week figures add the weeks of August 9 to 30, 2026, also on GPT-5.6 Terra.
- Counting unit. Each question was asked once a week, so “all four weeks” means four separate weekly answers to that question. The figures count single questions. The Segment Position tab groups questions by one condition at a time, so a question you win sits inside a card, and the question itself appears in that card’s filtered Prompt & AI Response.
- Outside the top 10. A brand that appeared in a category’s ranking in at least one of the four weeks and was never ranked 1st to 10th there. A brand in two categories counts once in each, so the 67 are 64 distinct brands. Brands named in answers but absent from the ranking in every week are not counted.
- Place in the answer. The order in which the answer names products, counting every product it names, including ones outside the category ranking.
- First three all four weeks. The brand’s place in that question’s answer was 1st to 3rd in each of the four weekly answers.
- The 16 pairs. Brand and question pairs where the brand was ranked below 20th overall in each of the eight weeks and was 1st to 5th in that question’s answer in each of them. 20 pairs met both conditions. In 16 of them, the passage around the brand’s first mention recommends the brand in each of the four answers from September 6 to 27. The other four were not confirmed as recommendations and were left out. The August answers were not read this way, so for those weeks the figures say only that the brand was in the first five.
- Products per answer. Distinct products counted as recommendations in the answer.
- The plain question. The panel asks no question without buyer conditions, so these figures carry no comparison with “what’s the best CRM?” style questions. That comparison comes from the Malthouse study above.