A budgeting app that shut down in 2024 is still recommended in 2026

Mint was the default budgeting app for a decade. Intuit closed it in 2024 and pushed users to Credit Karma. Anyone keeping up knows it is gone. The assistants, given live web search, mostly do not.

Across five budgeting questions, asked three times each, Mint shows up in Gemini's answers nine times out of fifteen. ChatGPT names it eight times. Both had web search on. Claude, also grounded, is the careful one at four. The cleanest are Google's AI Overviews and Mistral, which name it zero. Mistral is the sharp illustration: on my first run, without web search, it recommended Mint thirteen times out of fifteen, an answer straight from memory. Turn its search on and that drops to zero. The dead app was a thing the web could have told it about all along. See what grounding fixed.

Web search cuts the stale recommendations but does not remove them. The split inside Google is the tell: its chat app is the worst offender, its search overviews the cleanest.

Because that split is the interesting part. Google's AI Overviews, captured straight from the search page, did not recommend Mint once across the five questions. It named Mint only where the question named it first ("best Mint alternative"), and only as the thing to replace. Same company as Gemini, opposite result. The search surface is pinned harder to the live web than the chat app is, and it shows.

One honest caveat. One of the five questions is "best Mint alternative," which names Mint on purpose. That inflates the count. But Mint also turns up under "best budgeting app 2026," where nothing prompts it. The index now excludes an app from a question's counts when the question itself names it, so the ghost rate reflects real recommendations, not echoes.

The assistants do not agree

The more useful finding is that there is no single AI answer to chase. ChatGPT's top three are YNAB, Monarch and Rocket Money. Claude leads with PocketGuard and EveryDollar, and names YNAB less often, though when it does, it puts it first. Same questions, same week, different shelves. A brand that optimises for ChatGPT can be near-invisible in Claude. That gap is what a single-snapshot tool cannot show and a monthly index can.

What this is, and is not

A preview: five of thirty-eight budgeting prompts, US English. It is enough to show the method works and that the questions are worth asking every month, which was the point of running it by hand first. See the budgeting index for the full table, or how fitness compares.