When someone asks ChatGPT "what's the best app for tracking my habits?", it names a few products and moves on. Those few names get the user. If you build an app, it's worth understanding where those names come from, and what you can do to be one of them.
People already shop this way
In G2's March 2026 survey of 1,076 B2B software buyers, 51% said they now start their research in an AI chatbot rather than a search engine, and 71% rely on AI chatbots for software research. 69% ended up choosing a different vendor than they had planned because of the chatbot's guidance, and 85% said they think more highly of a vendor when a chatbot recommends it (G2, via Demand Gen Report).
Consumers are moving the same way. In US retail, visits from generative AI tools grew 693.4% year over year in the 2025 holiday season (Adobe).
Two ways an assistant knows about your app
- What it learned in training. Models are trained on a large snapshot of the web. If your app was written about before that snapshot, the model may know it. This changes slowly, only when new models come out.
- What it finds with live search. ChatGPT, Gemini, Claude and Perplexity can search the web while answering. They read the pages that come up for the question and summarize them. This can change within days of a new page being published.
The second one is where you have the most control, and it's how most "which app should I use?" questions get answered today.
Which sources they lean on
Pew Research found that Wikipedia, YouTube and Reddit are the most cited sources in Google's AI summaries, together making up 15% of everything cited (Pew Research Center). Beyond those, answers to "best app for X" questions tend to draw on the pages that rank for that question: roundups, comparison articles, review sites and directories.
In practice, an app that several independent pages mention is far easier for an assistant to recommend than an app that only describes itself on its own website. Your homepage tells the assistant what you are. Other pages tell it that you're worth naming.
How to get your app into the answers
- Find the real questions. Write down how a new user would describe their problem, not your brand name. "Best free invoicing app for freelancers" is a question an assistant answers. "AcmeInvoice review" isn't one most new users would ask.
- See who gets named, and from which pages. Ask the assistants yourself, or run a free AI visibility check. The full report lists every competitor named and every website the assistants cite.
- Get onto those pages. Ask the authors of roundups to consider your app, reply honestly in relevant Reddit threads, get reviewed on YouTube, and list your app in directories like Substitutr's.
- Make your own site easy to quote. One plain sentence saying what the app is and who it's for. A clear pricing page. An FAQ that answers the questions people actually ask. An llms.txt file with a short summary for AI crawlers.
- Check again over time. AI answers vary from one run to the next and change as new pages appear. A single check is a snapshot. Tracking the same questions over weeks shows whether your work is paying off.
Nobody can guarantee that ChatGPT will recommend your app, and you should be wary of anyone who promises it. But you can measure where you stand, see which sources the assistants trust, and work on showing up there.
Related: how people find apps in 2026, with the data on zero-click search and AI summaries.
