Google Backstory Tool: What It Can and Can't Check

Google Backstory checks for AI generation, edits and context

Google Backstory is an AI tool that investigates a single image. A user uploads the picture and types a simple question about it. Backstory then runs three checks on its own: whether the image is AI-generated, whether it shows signs of manipulation, and where else the image has appeared online.

The tool is built on Google's Gemini family of large language models. AI agents inside it decide which authentication tools to use and in what order.

The output is a written report with citations, together with a log of each step the tool took. The table below lists those three checks and the content-credentials check it also runs.

What Backstory checks for one uploaded image

Check	What it covers
AI generation	Whether the image is AI-generated
Manipulation	Whether the image shows signs of editing
Context	Where else the image has appeared online
Content credentials	Encrypted metadata added by companies that follow the C2PA standard

Content credentials follow the standard of C2PA, the Coalition for Content Provenance and Authenticity. Google, Adobe and the BBC are among the companies that signed onto its standard.

How India Today fact-checkers use Backstory for a first pass

India Today has a six-person fact-checking team. Its work might include an AI-generated archival photograph of a politician, a clip of deadly flooding taken out of its original context, and a deepfake of India's richest man. The team mainly uses Backstory to take the first pass on a piece of content, which speeds up the start of verification.

Before Backstory, a reporter might have run a picture through an AI detector, then a reverse image search, and finally built a timeline of its past appearances by hand. Backstory does all three at once.

Bal Krishna of India Today says the tool does the work of five different tools and five different logins in the same place. Mike Caulfield, who created the SIFT method, says Backstory shrinks an initial process of 50 minutes down to three.

Where Backstory falls short: look-alikes and unindexed posts

Backstory runs on large language models, and like most tools of that kind it cannot guarantee accuracy. A disclaimer appears at the bottom of the tool, similar to the one on Google's general AI assistants.

Caulfield names a common failure called conflation: the model mistakes one image for another that looks similar. Nieman Lab's reporter ran into it when asking for a report on the history of the reporter's own headshot. Backstory mistook it for the headshot of another journalist, Carlos Maza.

Coverage sets another limit. Its context tracing can reference only images that search engines have indexed. Caulfield points to TikTok, which is not indexed, and to closed groups such as WhatsApp or Signal: an image that started in either place may leave no trace.

Because of these shortcomings, India Today does not cite Backstory's reports directly in its published fact-checks. Reporters are expected to confirm by hand anything they pull from the tool, so Backstory speeds up the first pass while a person confirms what it produces.

Backstory is free but limited to Trusted Testers for now

As of the Nieman Lab report on 19 August 2026, Backstory is available only through Google's Trusted Testers Program. India Today's fact-checkers are among thousands of journalists, OSINT experts (investigators who work from public information), librarians and information literacy researchers whom Google has invited to try it.

The tool is free to use for now. Both facts come from that report, so access may have changed since. To keep a report like this as text in a verification log, the URL to Markdown tool converts a web page into markdown.