Why AI recommends your competitor's software

You built a great product and a sharp website, and ChatGPT still names the other guys. Annoying — but not mysterious. The model is reading rooms you've never decorated.

The shortlist is built from rooms you don't control

Ask ChatGPT for "the best [category] software" and it doesn't browse your homepage and rank it. It reaches for the sources it trusts to be objective — which, for software, means third parties. One analysis found that essentially every tool surfaced in ChatGPT's software answers had a Capterra presence, and nearly all had a G2 presence.¹ Not having a profile there isn't a disadvantage so much as a disqualification.

It gets more specific. Marketing pages rarely get cited; documentation, integration guides, and API references get cited roughly 3× more, because they read as objective fact rather than pitch.¹ And Perplexity, in particular, leans hard on Reddit — community discussion is a large share of what it cites.² So the honest ranking of what's deciding your software recommendations looks roughly like: review platforms, your docs, Reddit, third-party roundups… and somewhere down the list, the homepage you spent six months on.

Your comparison pages are quietly working for the other guy

Here's the one that stings. Most SaaS teams publish "[Us] vs [Competitor]" pages to win the comparison. To a language model, those pages are a tidy, structured summary of the competitor's positioning — often the cleanest one on the open web — sitting on a domain the model considers relevant to the query. So when a buyer asks the model to compare options, your page becomes a convenient source… for describing the competitor.

You're not imagining the recommendation going the wrong way. You may be feeding it. The fix isn't to delete the pages; it's to understand what the engines are actually extracting from them, which is exactly the kind of thing a diagnosis surfaces.

Why the engines can't agree on your category

If you've checked your category across a few assistants, you've seen them contradict each other. They will: across one analysis of hundreds of millions of citations, only about 11% of the domains ChatGPT cited were also cited by Perplexity.³ ChatGPT leans encyclopedic and recent; Perplexity leans Reddit and real-time; Google's AI Overviews lean on whatever already ranks. Three different source diets, three different shortlists.

Directional, from published source-mix analyses; proportions shift. The takeaway isn't the percentages — it's that there is no single "fix," only per-engine ones. The mechanic behind all of it is in why AI recommends some companies and ignores others.

The comparison page you built to win the deal is, to a model, the cleanest summary of the competitor's pitch on the open web.

What actually moves it

Not more homepage copy. The levers that matter for software are: a complete, current presence on the review platforms the engines treat as ground truth; documentation written to be lifted as factual answers; genuine third-party discussion (the unglamorous Reddit-and-community work); and a consistent, resolvable entity so the model is confident you're real. It's slower than a landing-page refresh and it's the work that actually changes the answer.

Before you spend a quarter on any of it, though, find out which lever is actually stuck — whether you're losing on review-platform absence, on docs the crawler can't parse, or on a comparison page feeding the competitor. Guessing is expensive.

← All insights

Guide · Published 2026-05-26 · [email protected]