An operations lead at a mid-sized firm asks an AI assistant: “what’s the best software for managing supplier contracts?” The answer names three products. A fourth company with a product that would suit them well isn’t mentioned, so it never makes the shortlist, and never finds out it was considered and passed over.
Software buyers now ask assistants questions like this before they visit a vendor’s site or fill in a form. If you aren’t in the answer, there’s nothing in your analytics to show it.
The test most teams run first
The usual first check is to ask an assistant “what is [our company]?”. It gives a reasonable description, and the team concludes it’s visible.
That test only shows the assistant knows who you are. The question that matters is the one that doesn’t name you: a buyer describing a problem and asking what to use. That’s where shortlists form, and a company can be described well by name and still be missing there. Our April 2026 study found exactly that for all seven companies it tested, though on a small sample and older models.
What to measure
Write prompts the way your buyers ask, in their own words, for each kind of buyer you sell to: the operations lead, the finance director, the IT team that has to approve it. Most should describe the need without naming any vendor. Then:
- Ask several assistants, several timesChatGPT, Claude, Gemini, Perplexity and Grok can all search the web while answering, and Google shows its own AI answers. They differ, so measure each separately.
- Count on the prompts that don’t name youComparison and brand prompts will nearly always mention you. Keep them apart, so they don’t flatter the rate.
- Put a range on every rateWork it out across prompts, not answers, and say how many prompts it rests on. A rate from five prompts is a rough guide; from 27 it’s firmer.
- Read how you’re describedWhen you are named, check the category, features and pricing the answer gives you. An answer that files you in the wrong category can do more harm than no mention.
The tracking guide walks through this step by step.
What software teams usually change
A clear page about the problem you solve
Answers to “what should I use for X” lean on pages that explain X well: what it is, how to choose, what to watch out for. If that page exists on other sites but not yours, the answer has little reason to name you. Write it in plain language, with facts you can stand behind.
One description, everywhere
Software companies with broad suites or a recent pivot often describe themselves differently on their site, in press coverage, on review sites and in marketplace listings. Assistants draw on all of these. Lead every key page and listing with the same plain statement of what you do and who it’s for.
The sites answers cite
Read which third-party pages the answers cite: review sites, comparison lists, analyst roundups, community threads. Where competitors appear and you don’t, that’s where to put effort, whether that’s inviting customers to review you or briefing the publications buyers read.
Measure again
Ask the same prompts after each change. If the ranges overlap, the change isn’t established yet. Keep measuring rather than declaring a win, and remember that even a clear change doesn’t prove your work caused it.
Measuring with Persipica
Persipica is built first for online retailers and consumer brands, and the measurement underneath is the same for any category: prompts written the way each kind of buyer would ask, five assistants and Google’s AI answers, three asks per prompt on the chat assistants, every answer and source kept, and a likely range on every rate. How Persipica measures sets it out.