Google pauses Ask Photos rollout over speed and quality issues

Google is hitting the brakes on another big AI rollout.

The search giant has quietly paused the expansion of Ask Photos, its Gemini-powered conversational search feature in Google Photos that first debuted at I/O last May.

The reason? It simply is not working well enough.

Responding to user complaints on X, Google Photos product manager Jamie Aspinall admitted that Ask Photos isn't where it needs to be, citing three glaring problems: latency, quality, and user experience.

Gemini gets a reality check

When Google first teased Ask Photos, the promise felt genuinely futuristic. Instead of typing rigid keywords, you could ask natural, context-heavy questions like what themes you used for a past birthday party or which national parks you visited.

Under the hood, Google deployed a specialized, custom version of its most capable Gemini multimodal models. The AI was supposed to inspect visual contents, read in-image text, and assemble tailored photo galleries alongside conversational summaries.

Instead, users were met with sluggish responses and missed queries. Aspinall confirmed that Google has now frozen the rollout at very small numbers.

He promised the team is working on an update due in about two weeks that aims to restore the speed and recall of the app's original search engine.

Meanwhile, Google also announced Tuesday that it is upgrading classic keyword search in Photos. Users can now use quotation marks to run exact text matches across filenames, camera hardware models, captions, and text inside pictures.

This is not Google's first emergency pause on an AI feature. Last May, the company pulled back AI Overviews in Google Search after viral embarrassments told users to put glue on pizza and falsely labeled Barack Obama as the first Muslim U.S. president. Last February, Gemini's image generation tool was famously paused within days after generating racially inaccurate historical depictions of the American Founding Fathers.

Why it matters

We think this pause highlights a hard truth for Big Tech: users do not want frontier AI if it breaks basic utility.

When someone searches their photo library, they are usually trying to pull up a receipt, a pet photo, or a vacation memory in seconds. In those moments, conversational summaries do not matter if the query takes five seconds to load and still misses the target photo. A snappy, deterministic keyword search will beat an unpredictable, sluggish LLM every single day.

For everyday users, there is no need to rush into testing AI search in Photos right now. The quiet winner here is actually the updated quotation mark search feature Google rolled out on Tuesday, which solves real search frustrations without the generative fluff.

Across the industry, this is another humbling reminder that cramming large models into everyday system utilities is fraught with UX hazards. Unless Google can make Gemini's multimodal reasoning nearly instant and rock-solid on recall, users will keep longing for the dumb, reliable search bars of the past.