a buyer asks which tool to use. does it say yours?
Software is bought by shortlist, and the shortlist increasingly comes from an assistant rather than a search page. Being absent from it is not a ranking problem. It is not being considered.
what your buyer actually types
These are the shape of the questions we ask on your behalf. You never write them: a set is drawn from a shared library for your category, and on the higher plans a set is written specifically for you.
What are the best project management tools for software teams?
Which CRM should a startup use?
What is the best alternative to Jira for a small engineering team?
Compare the leading customer support platforms for SaaS companies
Which analytics tool should I use if I already use Postgres?
Each question is asked repeatedly rather than once. Two identical runs minutes apart agree on only 36% of the brands they name, so a single answer is one draw and not a measurement.
what is actually at stake
The shortlist is the whole funnel
A category with a handful of credible names is decided at the shortlist, not the demo. Software buyers used to build that list from a search page and a peer; now they ask an assistant, and the assistant returns three names and a reason for each. If you are not one of the three, nothing downstream in your funnel ever sees that buyer.
You are compared whether you know it or not
Ask any assistant to compare tools in your category and it will do it, confidently, with reasons. It is drawing on whatever it read, which may be a review site's summary from two years ago, a Reddit thread, or a competitor's comparison page written specifically to beat you. That comparison is happening today and you have no copy of it.
Your competitor may have found this first
The pages an assistant reads to decide are ordinary pages: documentation, comparison posts, changelogs. The companies that noticed early have been writing for that reader for a while. This measures where you actually stand rather than where you assume you do.
the things people ask us before they start
- How do I find out if ChatGPT recommends my SaaS product?
- Ask it the question your buyer would ask, several times, and record what comes back. Asking once is not a measurement: identical runs of the same question minutes apart agree on only 36% of the brands they name, so a single answer is one draw. unbound geo asks each question ten times on the consumer surface and three times through the provider API, then reports how often you were named, who else was, and the exact text and cited pages behind each answer. The free scan does five questions on ChatGPT with no card.
- Why does ChatGPT recommend my competitor instead of us?
- Usually because of what it read, not because it judged you and decided against you. The answer cites pages, and those pages are the lever. unbound geo stores the full answer text and every source cited, so the question moves from "why not us" to "these four pages decided this answer and none of them are ours", which is something you can act on.
- Is AI visibility different from SEO for a software company?
- Related but not the same. You can rank first for your category and still be absent from the answer an assistant gives, because the assistant is summarising a handful of sources rather than listing ten links. Both are worth measuring, which is why the paid plans carry rank tracking, the answer box and AI answers together rather than selling them as separate products.
- Which plan should a SaaS company start on?
- Starter, for most. It covers 25 questions across ChatGPT and Gemini on a schedule, which is enough to cover a category and its adjacent searches. Growth adds Perplexity, custom question sets and daily rank tracking, and is the right step once you are acting on the findings weekly rather than reading them monthly.
The scan is free and needs no account. Most companies like yours start on Starter, at 49 USD a month, and the per-call provider costs behind every price are published.
This measures what the assistants say and shows you the evidence behind it. It does not promise that publishing anything will change the answer. Answers move for reasons no vendor controls, which is exactly why we sample them repeatedly rather than asking once.