Search Engine Roundtable Launches AI Tool to Detect New Search Features

By Tom Whitfield, SERP Journal

An AI tool that scans 23 years of archived search screenshots now answers a question SEO practitioners ask weekly: is this Google or Bing feature actually new, or have we seen it before? Barry Schwartz, founder of Search Engine Roundtable and CEO of RustyBrick, launched the free scanner at seroundtable.com/is-it-new on his long-running industry blog.

The launch itself is the confirmed fact here — everything about how the tool behaves comes directly from Schwartz's own description, not third-party testing.

What does the tool actually do?

The workflow is simple. A user uploads a screenshot of a search result element, adds descriptive text, and can highlight the specific portion in question — Schwartz compares the highlighting gesture to Google's Circle to Search. The AI then reads the image, classifies it, and determines matching criteria.

From there it scans every non-hero image (article body images, excluding large banner illustrations) across Schwartz's stories from the past 23 years and returns the closest matches. Users can also submit findings through a form for manual review.

The practical question for SEO teams and feature trackers: what does this change for verifying whether a SERP element (a search results page component) is genuinely new? It replaces the usual route — emailing Schwartz directly to ask — with a self-service lookup against one of the largest independent archives of search interface screenshots in existence.

How it was built

Schwartz credits two RustyBrick developers. He said: "I was joking around with the RustyBrick team about this idea and two of the RB developers were like, sure we can build that. So I said, go ahead."

The backend works in three stages:

• A preprocessing run where AI indexed all historical images on the site to speed up and cheapen matching
• A daily indexing job that adds images from new stories
• Multiple AI models running in parallel, with Schwartz able to toggle between them to compare performance

Uploaded images are stored on the backend for debugging.

Why there is a paywall

The tool costs roughly 7 cents per scan to run. To keep bots from burning through that budget, Schwartz added a light gate: users enter an email, receive a code, and are subscribed to his newsletter — with the option to unsubscribe at any time. Daily scan limits apply per user, and the tool has an overall daily cap that displays a notice when reached.

Schwartz addressed the tool's maturity directly: "It is not perfect but it is a pretty cool concept."

Tool-signal versus confirmed behavior

It is worth separating what is verified from what remains to be seen. Confirmed: the tool exists, it runs on a pre-built image index refreshed daily, and it uses multiple AI models that Schwartz can switch between. Unverified: its accuracy rate. Schwartz says the team did "a few weeks of testing" and that he expects continued tweaks once real-world usage reveals weak points — but no accuracy figures, match-rate statistics or benchmark numbers have been published. Treat the results it returns as a lead to investigate, not a verdict.

Who benefits most

The natural audience is narrow but demanding: SERP-feature trackers, SEO news desks and in-house teams documenting interface changes across Google and Bing. For publishers covering search, it offers a faster first pass on novelty claims before an editor asks whether a "new" feature actually dates back to 2016. A more technical writeup is available on the RustyBrick blog, and Schwartz has opened an X forum thread for feedback.

What to monitor next: whether Schwartz publishes accuracy data or model-comparison results after real-world usage accumulates, and whether the daily index keeps pace with the volume of screenshots his coverage generates.