You typed your category's buying question into ChatGPT, and it named three companies. None of them was you. One of them is a competitor you beat on every deal you're both in.
That's not a ranking problem and it isn't bad luck. Answer engines assemble a response from sources they can retrieve, quote, and verify — so a competitor being named means their footprint clears those three bars and yours doesn't, somewhere specific. In practice it's almost always one of five asymmetries, and the diagnosis matters because they differ enormously in cost to fix.
1. They answer the question; you describe your services
The most common cause, and the least expensive to fix. Your competitor has a page whose entire job is answering *how much does X cost* or *best X for Y*. You have a services page that describes your capabilities beautifully and answers no question anyone typed.
Models retrieve against intent. A page titled *Our Approach* does not match *what should I look for in an X vendor*, no matter how good the approach is. The fix is boring and mechanical: one page per real buying question, with the answer in the first two sentences.
2. They publish facts; you publish adjectives
*Industry-leading, tailored, end-to-end, best-in-class* — none of it can be quoted, because quoting it would commit the model to a claim it can't support. A model reaches for what it can defend. A sentence like *a production agent is a flat fee from $8,000, deployed in about 90 days* survives extraction. *We deliver bespoke AI solutions that drive transformative outcomes* does not.
This is why published pricing is such an outsized advantage — almost nobody does it, and it's the single most quotable fact in a vendor category. The Princeton GEO study found that adding citations, quotations, and statistics lifted visibility in generative-engine answers by up to 40% (Aggarwal et al., KDD 2024). Specificity is the mechanism.
3. Someone else vouched for them
Answer engines weight third-party corroboration heavily, because a company describing itself is the weakest possible evidence. If your competitor appears in roundups, comparison posts, directories, podcasts, and reviews — and you appear only on your own domain — the model has one source for you and eight for them. It will name the one it can cross-check.
This is the slowest of the five to fix, and the reason to start now rather than next year.
4. The engine isn't sure you're one company
Entity confusion is invisible from the inside and lethal. Your site says *Acme AI*, LinkedIn says *Acme Technologies Inc.*, the directory listing has a defunct address, and an old subsidiary name still ranks. A model that can't resolve those into a single confident entity will quietly avoid naming you at all — being wrong is worse for it than being unhelpful.
Consistent naming, an address that matches everywhere, and organization schema that states plainly who you are and what you sell resolves this. It's unglamorous and it moves the number.
5. You're checking a query nobody asks
Worth ruling out before you spend anything. If your test query was *best forward-deployed AI engineering firm for mid-market healthcare*, an absent answer may just mean nobody phrases it that way. Test the language your buyers actually use — and don't test your own brand name, which measures recognition rather than acquisition.
How to tell which one you've got
- Named competitor, you absent, on a question you *do* have a page for → #2 or #4. Your page exists but isn't quotable or you aren't verifiable.
- Named competitor, and you have no page for that question → #1. Cheapest fix on the list.
- You appear for your brand name but never for category questions → #1 plus #3.
- Inconsistent — cited in one assistant, invisible in another → #3 or #4. Different engines lean on different indexes, so a thin external footprint shows up unevenly.
The fastest way through this is to stop guessing. The free citation audit runs your category's real buying questions across the major assistants and shows where you're absent, who's being named instead, and which of the five gaps is actually yours. If you want the mechanics underneath first, how assistants choose who to recommend covers retrieval, extraction, and verification, and Ascent is the program that closes the gaps in order of cheapest-first.
One consolation: your competitor almost certainly didn't earn this deliberately. Most categories still have no serious GEO players, so the company being cited today is usually just the one that happened to publish specifics. That's a lead measured in months, not years — and it's reversible.
