Somewhere right now a buyer is typing “best alternatives to” followed by your biggest competitor’s name into ChatGPT. That buyer is the most valuable person in your market. They have a budget, they have the problem your category solves, and they have already decided to leave the incumbent. The only question left is who they leave for. They are describing your competitor because they do not yet know to describe you.
That single query, the comparison query, is where the most winnable money in AI search sits. And almost nobody treats it as its own game.
Why the comparison query is a different game
Most advice about AI search treats getting cited as one undifferentiated goal. Be clear, be specific, earn trust, show up. That is true at the level of the whole site, but it hides something important: the engine does a different job for different questions, and the comparison question is the one with a buyer’s hand already on their wallet.
When someone asks an open category question, “what is the best issue tracker,” the engine is doing discovery. It returns a broad shortlist and the buyer is still browsing. When someone asks a comparison question, “best alternative to Jira for a small team” or “Linear vs Jira for startups,” the engine is doing something narrower. It is adjudicating. The buyer has a reference point and wants a verdict relative to it. There is far less ambiguity about intent, which is exactly why it converts.
I wrote about the broad version of this in how B2B buyers choose software with AI. The comparison query is the bottom of that same funnel, the moment right before the demo form, and it is governed by its own rules.
The counterintuitive part: you win by conceding
Here is the move that feels wrong and works anyway. To win the comparison query, your source has to admit where you lose.
When the engine assembles a comparison answer, it is looking for a source that reads like a neutral arbiter, not a sales page. It rewards the page that says “use Competitor X if you need deep enterprise permissioning, use us if you want something a five-person team can run on day one.” That page gets pulled into the answer as the even-handed reference. The page that says “we are the best choice for every team of every size” gets treated as marketing and skipped, because the engine has learned that self-praise carries no information.
So the comparison query inverts the usual instinct. Everywhere else, marketing wants to minimize the competitor and maximize itself. In the place where it matters most, the honest comparison is the one that gets cited, and being generous about where a competitor genuinely wins is what earns you the citation on the half where you genuinely win. The engine is not grading your loyalty to your own product. It is grading whether your page is useful to a buyer trying to choose, and a buyer trying to choose needs to know the tradeoffs.
This connects to a principle I keep coming back to: whoever the engine cites writes your label. If you refuse to write an honest comparison, someone else writes it, and they have no reason to be kind. The fix for a bad label lives at the source layer, and on the comparison query you have an unusual amount of control over that source, because you are allowed to be one of the sources.
What I got wrong
The first comparison page I helped build did the obvious thing. It listed our strengths against the incumbent in a tidy table, every row a win for us. It was accurate, it was well written, and the engine never touched it. Months of it sitting there, cited zero times in the comparison answers it was built for.
We rewrote it to concede. We added the two categories where the incumbent was genuinely better and said so plainly, with the specific use case where a buyer should pick them instead. Within a recrawl cycle the engine started pulling from it, and the lines it pulled were our real advantages, the ones now made credible by sitting next to honest losses. The version that admitted weakness got cited for its strengths. The version that claimed everything got cited for nothing.
The lesson held up every time after: the engine does not reward the page that wins every row. It rewards the page a buyer would trust, and nobody trusts a comparison where one side wins everything.
How to actually compete for it
The comparison query rewards a specific kind of asset, so build that asset deliberately.
Start by finding the exact comparisons your buyers run. Not the ones you wish they ran. Run “best alternative to [each major competitor]” and “[you] vs [each competitor]” and “best [category] for [your strongest segment]” through the engines yourself and write down who the answer names and which sources it cites. That is the same query-mapping discipline from the buyer query map, pointed at the bottom of the funnel.
Then publish the honest comparison for the matchups where you have a real, defensible win for a specific buyer. One page per serious competitor, written as a buyer’s decision aid, not a brochure. State who should pick them. State who should pick you. Be specific about the segment, because “best for startups” and “best for a regulated enterprise” are different verdicts and the engine increasingly answers the qualified version.
Last, make sure the third-party sources tell the same story, because the engine cross-checks. If every review site and forum thread frames you one way and your comparison page claims another, your page loses the tie. Earning that outside corroboration is slow and cannot be faked, which I covered in the source-layer fixes, but on a comparison query it is the difference between being the cited arbiter and being ignored.
Why this is the winnable one
A two-month-old site is not going to beat an established player on the broad head term. The authority gap is too wide and the query is too contested. But the comparison query is narrower, less crowded, and turns on a quality most incumbents will not produce: a genuinely fair account of their own weaknesses. Big companies almost never publish “use our competitor if you need X.” That refusal is your opening.
The buyer searching for your competitor’s alternatives is not loyal to your competitor. They are already leaving. The only thing standing between them and you is whether the engine has an honest source to point them at, and whether that source is yours. If you want to see which comparison queries already mention you and which ones do not, that is one of the first things a paid audit maps, and the full method is on the methodology page.