On this page
If you run a young software company, a buyer asking an assistant for the best tool in your category and a buyer asking the same question with their own situation added get different lists. They are two separate doors, not two rungs of one ladder, and different people own them.
In our August research across three young software categories, 75.84% of the best-tool answers named at least one young product (founded in 2023 or later) that sits on the list most of the time. When the buyer added their situation, 38.89% did.
Two Questions, Two Owners
Every category has a broad question: the buyer asking which tool is best for the job, with nothing about their own situation. Buyers word it many ways, and the answers keep returning to the same small pool of products.
Most of the products named in three or more of those answers are not new companies: 10.68% were founded in 2023 or later. The best-tool list has few young companies, but those holding a place on it get named often.
Now add the buyer's situation: a six-person sales team that lives in video calls, interviews with early customers, a support agent that hands over to a person when it gets stuck. The answers change shape.
In the same number of answers, the questions with a situation brought up 1.86 times as many different products. Across three ChatGPT askings of the same question, 19.69% of the products appeared every time on average, and 59.19% appeared once.
That's the opening for a new product: one that fits the situation can get named without being on the best-tool list. A single mention proves little, though.
Piotr BabelPartner @ QveryWe ask each situation question three times on each engine, on different days, and count how many of the three answers name you, because one answer misleads: on narrow questions, 59.19% of the products ChatGPT named showed up in only one of three askings. Keep that count per question every month, and set it beside the pages each answer linked, so you see which of your pages the engines read when they name you.
Narrow Answers Kept Fewer Products From One ChatGPT Asking to the Next
% of the products named for a question
View the data
| Label | Broad question | Narrow question |
|---|---|---|
| Named in all three askings | 35.49% | 19.69% |
| Named in one asking only | 43.90% | 59.19% |
What the Answers Read
The pages behind the two kinds of answer differ as well, and that suggests who in your company owns each one.
On ChatGPT, 69.46% of the pages linked as sources under best-tool answers rank or compare tools, judged by the address alone. Under the answers about a situation, 27.61% sat on the website of a product the same answer named, far more than under best-tool answers.
The best-tool list is a partnerships job: getting a new product onto the pages that rank tools each engine reads, by means their editors accept.
The situation answer is product marketing’s job: a page on your own site for each situation your product fits, in the words a buyer uses, with the facts an answer can repeat.
Vlad ShvetsPartner @ QveryFor a young product, put the first money into your own situation pages before you chase a place in the roundups. Under answers about a buyer's situation, 27.61% of the linked pages sat on a named product's own site, and those pages are yours to write this month; the best-tool list rests on ranking pages you can only pitch. Give product marketing the situation pages now, and fund partnerships once most situation questions name you.

Which door you are behind shows only in who each answer names, asking after asking.
What Qvery’s Agents Do With Your Product Facts
Our agents tell you whether your next page should explain a buyer’s situation or whether your product needs a place on the pages that rank tools. Each month they report, for every situation question, how many of three askings named you.
Zero suggests starting with a page of your own for that situation. One of three means you are one name among many, and that page needs the facts an answer can repeat. Two or three on most questions is when we would make the best-tool list, and the pages behind it, next quarter’s work.
The agents keep your product facts current on your pages and draft notes to the editors of the pages that rank tools; a person approves each page change and sends each note.
Every page change and editor note they draft is logged, approved or not.
The calculator prices that page and editor work for a software company of about 200 people: a product marketer and a partnerships lead, each spending about a quarter of the year on it. At about $127,800 a year each, the Bureau of Labor Statistics’ mean pay for marketing specialists with benefits, that is about $63,900 a year.
With the agents taking about 55% of it, at about $400 a month to run, it costs about $33,555, saving about $30,345 a year and 572 hours; a $10,000 build is paid back by about the fourth month. Each figure is our assumption, to replace with your own.
Product Page and Roundup Upkeep ROI Calculator
What Keeping Product Facts Current On Your Pages And Drafting Notes To Roundup Editors Costs You, And What An Agent Saves
Saved Per Year
$30,345
572 hours handed back, after $4,800 a year to run it
- Hours Handed Back
- 572 hours a year
- Agent Running Cost (Estimate)
- $400 a month
- Payback Per $10,000 Of Build
- 4 months
On these numbers, the agent saves $30,345 a year after its own running cost, and every $10,000 of the build is earned back in 4 months.
Behind the report, the GEO Agent prepares questions from your software category, brand and competing products, then runs them daily on ChatGPT and Google AI Mode in the countries you sell in. Its records show how often you are named beside competing products, by topic and country.
We go through where the recommendations came from with your team, then improve the pages and check their facts. Each week we count the hours of page and editor work the agents took off your product marketer and partnerships lead.
The First Two Weeks
We begin from your product facts, last quarter’s pages and your list of buyer questions, and within two weeks of access the agents are reading your buyers’ answers and drafting the first page changes.
Frequently asked questions
How long does it take a new SaaS product to show up in ChatGPT's recommendations?
One month of answers can't establish a timeline. Our August 2026 research found young companies on the broad shortlist, while narrow questions admitted more products but reshuffled them between askings. Measure repeated appearances before treating a mention as a foothold.
Should a new SaaS product target the broad best-tool question or narrow questions first?
Measure both, then fund by the reading. Answers about a buyer's specific situation linked the named products' own sites far more often than best-tool answers did, so your own situation pages come first. Once most askings name you, the pages that rank tools behind the best-tool list become the work.
Why does ChatGPT recommend different tools each time I ask the same question?
Its lists change between askings, and more on specific questions. Asked the same specific question three times, 19.69% of the products ChatGPT named appeared in all three answers on average, and 59.19% in only one; best-tool questions kept more of their names. Read your place as a share of askings, never from one answer.
Which pages do AI answers cite when they recommend software?
It depends on the question. On ChatGPT, 69.46% of the pages linked under best-tool answers rank or compare tools, judged by the address. Under answers about a specific situation, 27.61% sat on the website of a product the answer named.
Does Google AI Mode cite the same pages as ChatGPT?
Not the same ones. Its best-tool answers also leaned on pages that rank or compare tools, but it linked none of the four small ranking sites that ChatGPT linked often in these categories. Plan the pages behind the list from both engines' sources.
How do I know whether a customer found us through an AI answer?
The answers can't tell you. Ask every new prospect where they first heard your name, and keep that count beside the answer records. It is self-reported, and still the closest line from an answer to a deal you can draw.
Who should own broad and narrow AI recommendation questions?
Give partnerships responsibility for the roundups and rankings that broad answers cite. Product marketing owns the pages explaining the situations your software fits; keep the two readings separate so the founder can decide which job needs attention.
What should I do if AI answers never name my product?
Ask each narrow question three times on each engine, on different days. If none of the answers names your product, write a page explaining that situation and the facts about how your product handles it before pursuing the broad shortlist.
When should I put effort into software roundups?
If most narrow questions name your product in two or three of three askings, we would make the broad list next quarter's work. Read the roundups each engine cites and approach their editors through the methods they accept.
Does a citation prove a page caused an AI recommendation?
No, a citation shows which page the answer pointed to, not why the engine selected a product. Our suggested order of page work and roundup work is judgment based on the observed split between broad and narrow answers.
How many young companies make ChatGPT's best software lists?
Few. Of the products named in three or more best-tool answers across three young software categories in our August 2026 research, 10.68% were founded in 2023 or later. Those few hold their place: 75.84% of the best-tool answers named at least one of them.
Which decisions stay with a person?
A person approves every page change and sends every note to an editor, and the founder chooses which door to work on next. The agents draft; they publish and send nothing themselves.
Do AI answers name more products when the buyer describes their situation?
Yes. In the same number of answers, questions with the buyer's situation named 1.86 times as many different products, and fewer of them came back from one asking to the next, so a small product can get named there before it makes the best-tool list.