Your Users Are Talking to AI About Your App. Are You in the Answer?

Somebody out there is typing "best free app to track lifting and running" into a chatbot right now. The chatbot will name a few apps. I want to know if mine is one of them. The Shift I Can't Ignore I'm Bud, an indie dev at Noland Software. My app is Power Leveling, a free fitness logger on Google Play for lifting, running and intervals. It gives you a Power Level score that climbs as high as 30,000, has a built in Tabata timer, and a friends feed so you can see who actually showed up today. For years, getting found meant two things: app store search and Google. That's still true. But a third door has opened, and it's getting busy fast. At OpenAI's DevDay in October 2025, Sam Altman said more than 800 million people use ChatGPT every week. Similarweb estimated that AI platforms sent over 1.13 billion referral visits to the top 1,000 websites in June 2025, up 357% from a year earlier. Statcounter's data from May 2025 put ChatGPT at 79.8% of chatbot referrals, with Perplexity at 11.8%, Microsoft Copilot at 5.2% and Gemini at 2%. Keep that in perspective, though. In the same Similarweb report, Google search sent 191 billion referrals in June 2025. AI is a small slice of traffic. It's just the slice that's growing. How a Chatbot Decides What to Recommend As far as I can tell, there are two main ways an app ends up in an AI answer. Training data. The model learned about your app from the web text it was trained on. If nobody wrote about you before the cutoff, the model may simply not know you exist. Live retrieval. Many assistants now search the web while answering, then summarize and cite what they find. OpenAI documents a dedicated search crawler and says sites that opt out of it won't be shown in ChatGPT search answers. Google says AI Overviews and AI Mode may use a query fan out technique, running several related searches to build one response. The second path is the one an indie dev can actually influence this month. If the assistant is searching, it needs to find pages that clearly say what your app is and who it's for. GEO Is Real Research, Not Just a Buzzword The term Generative Engine Optimization comes from a paper by researchers at Princeton, IIT Delhi and independent collaborators, published at KDD 2024. They built a benchmark of 10,000 queries and tested ways of rewriting content to get it surfaced in generative answers. Their finding: GEO methods boosted visibility by up to 40% in generative engine responses, and up to 37% on Perplexity. Adding citations, quotations from relevant sources and statistics helped a lot. Keyword stuffing didn't. They also found results varied by domain, so there's no universal trick. You'll also hear AEO, answer engine optimization. Same basic idea: write content an AI can confidently quote. What I'm Doing for Power Leveling Here's my checklist. None of it is magic, and most of it is just good SEO wearing a new jacket. Plain descriptions on my site. nolandsoftware.web.app should say in one sentence what Power Leveling is: a free fitness logger for lifting, running and intervals with a Power Level score up to 30,000. No clever taglines that a model has to decode. Consistent naming. On Google Play it's "Power Leveling by Noland." Power leveling is also a gaming term, so I need the app name, the studio name and the fitness context to show up together everywhere. Otherwise a model could easily mix me up with video game boosting services. Structured data. Schema markup describing the app, its price and its category. Google is clear that you don't need special schema for AI features, but it does say your structured data should match the visible text on the page. An FAQ page. Real questions people ask: Is it free? Does it track running and lifting together? How is the Power Level score calculated? Is there an interval or Tabata timer? Those are the exact shapes of questions people type into chatbots. Getting mentioned elsewhere. Review sites, fitness blogs, YouTube creators, app roundups, small press. The GEO paper suggests sources and citations matter, and an assistant searching the web will lean on what other people say about you, not just what you say about yourself. Letting the crawlers in. Checking my robots.txt so search crawlers aren't blocked by accident. llms.txt, maybe. It's a proposal from Jeremy Howard, first published in September 2024: a markdown file at your site root that gives AI tools a clean summary and links. I'll probably add one because it takes ten minutes. But I'm not counting on it (more on that below). The Test I'm Running Here's the honest part. I'm going to ask ChatGPT, Gemini, Perplexity and Google's AI Overviews the questions my ideal user would ask. Things like "fitness app with a strength score," "free app that logs lifting and running," and "workout tracker with a Tabata timer and friends feed." I'm going to check three things. Does Power Leveling show up at all? If it does, is the description accurate? And which sources get cited, because those are the places I need to be. I haven't got results to share yet. My honest guess is that a small indie app with a fresh website won't show up much at first. That's kind of the point. I want a baseline, then I want to see if the checklist above moves it. The Counterpoint: Don't Bet the Farm on This I'd be lying if I said this is a sure thing. It's unproven for small apps. The GEO paper is a solid study on a benchmark, not a promise about your app on a given Tuesday. It's volatile. Models update, retrieval changes, and the same prompt can return different apps on different days. Google itself notes that AI Mode and AI Overviews may use different models, so their answers and links vary. The zero click problem. Pew Research analyzed March 2025 browsing data from 900 US adults. When a Google AI summary appeared, people clicked a traditional result in 8% of visits versus 15% without one. They clicked a link inside the summary in just 1% of visits. Being cited isn't the same as being visited. AI can get your app wrong. Models hallucinate. A chatbot could invent a price, a feature, or confuse Power Leveling with a gaming service. Part of this work is defense: making the true facts easy to find so the wrong ones have less room. llms.txt may do nothing. Google's John Mueller wrote in June 2025 that "no AI system currently uses llms.txt," and also compared it to the old keywords meta tag. Google's own docs say you don't need new machine readable files or AI text files to appear in its AI features. So my plan is simple. Build the clear, honest website I should have built anyway. Earn real mentions. Measure. Don't chase hacks. Your Turn Have you asked a chatbot to recommend an app like yours? Did you show up, and was what it said about you actually true?

Originally posted on X: https://x.com/ArgileNoland/status/2113784090344276172