Short answer: open the photo in the Google Photos app, tap More, then About, and scroll to “How this was made”. If the file carries Content Credentials you will see lines such as “Media captured with a camera” or “Edited with AI tools”. If the section is missing, the file carries no signed history, and that absence proves nothing on its own.
Someone sends you a photo. It looks a little too clean, or the story attached to it feels off, and you want to know whether AI touched it. The Google Photos how this was made panel is the quickest check most people already have on their phone, because it sits inside an app most Android phones ship with and many iPhone owners install too. It reads a signed record inside the file instead of guessing from pixels, which is why it can settle some questions in seconds. It also has hard limits, and this guide works through them with two real photos checked on 14 November 2026.
Checked 14 November 2026. Google’s help pages are the source for every panel label quoted here, and the two worked photos were inspected file by file on the same day.
What is the “How this was made” panel?
In short: it is a brief summary of a photo’s signed history, shown in the About panel of the Google Photos app.
Content Credentials are a signed record of how a media file was made. The standard behind them is C2PA, built by the Coalition for Content Provenance and Authenticity, and Google Photos is one of the apps that can read it. When a file carries that record, the Google Photos app sums it up in a section called “How this was made”. The section can show up to four categories, and each one names the company that signed the information, for example “Info by Google LLC”. When more than one company has signed, the most recent signer is listed first.
Think of it as a logbook rather than a verdict. It tells you which tools touched the file and in what order. It does not tell you whether the photo is real or fake, and Google says exactly that on its own help page. A real photo can carry a long logbook, and a fully generated image can carry none at all if the file was saved by an app that never wrote the record. That distinction runs through everything that follows. The panel is a starting point, not a final answer.
How do I open the Google Photos how this was made panel?
In short: open the photo in the Google Photos app, tap More, then About, and scroll to “How this was made”.
Google’s own steps are short, and they are the same on Android and iOS:
- Open the Google Photos app on your phone.
- Open the photo you want to check. If someone sent it to you in a chat, save it to your phone first, then open that saved copy in Google Photos.
- Tap More, then About. On Android you can also swipe up on the photo to reach the same details screen.
- Scroll down to “How this was made”. If the file carries Content Credentials, the section is there. If it does not, the section is missing entirely, and there is nothing to tap.
Two points matter before you start. First, the panel is app only. Google’s help page states plainly that Content Credentials are not available on Google Photos web, so checking the same photo in a desktop browser will show you nothing, even when the phone app shows a full history. Second, you need Android 8.0 or above, and the photo needs to be a file the app can read. Google first announced the AI info section in October 2024, and the panel has been reaching more phones since then. By November 2026 it is live in the app on both Android and iOS, but it still only appears for files that carry a compatible record, and it is still not on the web.
If the photo came to you inside WhatsApp or another chat app, save the original file rather than a screenshot of it. A screenshot is a new photo of your screen. It carries none of the first photo’s history, so the panel will be empty even when the original was fully documented. The same applies to a photo you re-saved from a social feed. Always check the file itself, never a picture of the file.
What does each line in the panel mean?
In short: each line describes one part of the file’s history, where it came from, whether AI edited it, or whether the edits were ordinary ones.
| What the panel says | What it means | What it does not mean |
|---|---|---|
| “Media captured with a camera” | The file’s record says a camera took this photo. Google’s example is a default Pixel 10 capture. | It does not prove the scene is genuine. A camera photo can still be staged. |
| “Edited with AI tools” | Generative AI changed the photo and created new pixels, for example with Magic Eraser. | It does not mean the whole photo is fake. One erased object can trigger it. |
| “May have been edited by AI tools” | Another app recorded an edit but did not say whether AI was used. | It is a maybe, not a finding. Treat it as a prompt to check further. |
| “Edited with non-AI tools” | Ordinary edits were recorded, such as crop or rotate. | It does not rule out AI. An AI edit made by an app that writes no record will not appear. |
| No section at all | The file carries no Content Credentials the app can read. | It proves nothing about the photo. Most photos online carry no record. |
Read the lines narrowly. “Edited with AI tools” records that an AI tool was used. It does not say how much of the photo changed. Magic Eraser removing one stranger from the background and Magic Editor rebuilding half the frame can both produce the same line. If the line points at Google’s editor, our guide to telling if a photo was edited with Google Magic Editor walks through the three checks that pin it down. The panel tells you that AI was involved. It leaves the scale of the change for you to judge by looking at the photo.
“May have been edited by AI tools” is the line people misread most. It appears when a third-party app recorded an edit in the file’s history but did not record whether AI powered it. That is an honest shrug from the file, not a quiet accusation. Treat it as a reason to run the other checks in this guide, not as proof of anything.
Also note the signer line. “Info by Google LLC” means Google signed that entry. A file can carry entries from more than one company, and Google lists the most recent signer first. If you see a signer you do not recognise, pause and check the file another way. A strange signer is not proof of fakery, but it is a reason not to take the photo at face value.
What is the difference between “How this was made” and “AI info”?
In short: “How this was made” reads a signed C2PA record. “AI info” reads older IPTC metadata, which anyone can edit or strip.
Google Photos actually has two routes to the same question, and they are not equal. The newer one is “How this was made”, which reads Content Credentials, the signed C2PA record. The older one is “AI info”, which reads IPTC metadata. Google added the “AI info” section in October 2024 for photos edited with Magic Editor, Magic Eraser and Zoom Enhance, and for composite photos made with Best Take and Add Me, which combine elements from several shots without using generative AI.
IPTC is a long-standing standard that photographers and news organisations use to describe images, and it works well for captions and credits. For AI history it has one weakness, and Google’s own help page states it twice: IPTC metadata can be edited or removed. The “AI info” section is also removed entirely for photos that have not been edited with AI or that do not follow IPTC standards, so its absence tells you nothing.
So a photo can show one section, the other, or neither. A signed C2PA record is harder to alter quietly, because each editor signs its own entry and tampering shows up as an error. An IPTC field is plain text inside the file, and any editor can rewrite it. The practical rule is simple: trust “How this was made” more than “AI info”, and trust neither as proof that a photo is untouched. If you want the fuller version of the signed-record route, including how to read it on the web, our Content Credentials guide covers it step by step.
Worked example: two real photos checked on 14 November 2026
In short: a real camera photo showed no panel at all, and a real C2PA test file showed a signed history of ordinary edits. The panel reports what the file carries, nothing more.
To show what the panel can and cannot do, I checked two real files on 14 November 2026, reading their bytes directly rather than guessing. The panel itself only exists inside the Google Photos app, so the panel wording quoted in this article comes from Google’s help page, checked the same day. The file facts below are what I measured.
Photo 1: an ordinary camera photo
The first file is a real photograph of the Sydney Opera House, published on Wikimedia Commons, taken with a Canon EOS 5D Mark IV. The file is 10,815,149 bytes and 6,479 by 3,644 pixels. Its EXIF data is intact: Canon as the maker, Canon EOS 5D Mark IV as the model, and Adobe Photoshop Lightroom Classic 12.4 listed as the software that processed it. The file contains an XMP packet, which is where IPTC data would live, but it carries no IPTC DigitalSourceType field at all. More importantly, it contains zero bytes of C2PA data: no JUMBF box, no APP11 segment, no manifest of any kind.
Open this file in Google Photos and there is no “How this was made” section and no “AI info” section either. Nothing is wrong with the photo. It is a genuine camera shot with full EXIF data. The panel is empty simply because no app ever wrote a signed history into this file. Lightroom processed it, the photographer exported it, and none of those steps added Content Credentials. That is the normal state of most photos on the internet, and it is the first thing to understand about this panel: an empty result is the default, not a finding.
Photo 2: a file with a real C2PA record
The second file is CA.jpg, a test image from the C2PA reference project. It is 166,864 bytes and 1,024 by 683 pixels, and it carries a genuine C2PA manifest store: two JUMBF segments in the APP11 marker, 64,010 and 53,259 bytes, holding a claim, assertions, an ingredient list, a signature and thumbnails. The claim generator is recorded as make_test_images with c2pa-rs version 0.33.1, and the manifest is signed with the C2PA test certificate chain, including a “C2PA Test Signing Cert” and a “FOR TESTING ONLY” root. The recorded actions are “opened” and “colour adjustments”, and the file lists one ingredient, which is the source image it was built from.
For this file, Google Photos has something to show. The recorded edits are ordinary ones, so the honest reading is the non-AI line, not an AI flag. Two caveats matter. First, this manifest is signed with a test certificate rather than a production signature from a camera maker or an app, so a phone may show the history differently, or show an “Unrecognized” style error, instead of the clean summary above. Second, the actions recorded here are mild: opening a file and adjusting its colours. A file with a short, boring history is exactly what a genuine workflow looks like. A long history is not suspicious on its own, and a short one is not proof of innocence.
The contrast between the two files is the lesson. The camera photo is real and shows nothing. The test file is a lab sample and shows a full signed history. The panel reports what the file carries. It never reports what the file does not carry, and it never tells you whether the scene in front of you actually happened. Adobe Firefly is a useful third case to know about: Firefly signs its output with Content Credentials, so a Firefly image opened in Google Photos can show its origin directly. Our guide to spotting Firefly images works through a real Firefly file, manifest and all.
What does an empty panel prove?
In short: nothing. Most photos carry no signed history, and the apps that pass photos around strip the record away.
An empty panel has at least five ordinary explanations, and only one of them involves deception. The photo may predate the C2PA standard entirely. It may come from a camera or an app that never writes Content Credentials. It may have passed through WhatsApp, Messenger or another chat app that recompresses the file and drops the metadata. It may be a screenshot, which is a brand new photo with no history of its own. Or someone may have edited the metadata out, which is easy for the IPTC route and possible for the C2PA route by re-saving the pixels as a fresh file.
“No label means real” is the exact mistake this panel invites, and it is the one to guard against. A scammer does not need to fake a signed history. They only need a photo with no history at all, which is what most photos look like. An empty panel should leave you exactly where you started: with a photo you have not verified yet.
When the panel is empty, fall back to the checks that do not need metadata. Reverse search the photo to see where else it appears. Look at the source: who sent it, and what do they gain if you believe it? Look at the pixels for the usual tells, warped hands, mismatched light, text that dissolves into nonsense. And if the photo matters, ask for the original file. A sender who has it will usually share it. A sender who cannot produce it has told you something too.
What if the panel shows an error?
In short: Google names three errors, and each one points at a different problem with the file’s record.
Sometimes the section appears but carries an error message instead of a summary. Google’s help page lists three, and they are worth knowing apart.
- Incorrectly modified. Content Credentials are built on a history that each editor signs in turn. If that history was tampered with, or changed without a proper update to the C2PA data, the panel reports the record as incorrectly modified.
- Incompatible or unsupported. Google Photos supports Content Credentials versions 2.1 and 2.2. Some third-party apps store the metadata remotely, separate from the photo itself, and Google Photos does not support remote metadata, so those files show this error.
- Unrecognized. The file contains malformed Content Credentials, often because the app that made the file had a bug. Google Photos can see that metadata is present but cannot read it.
Google adds one more caveat: in offline mode, Content Credentials are not always signed or maintained. An error is not proof of fakery. It means the record cannot be trusted as written. Treat the photo as unverified and use the other checks below, just as you would for an empty panel.
What should I check when the panel is not enough?
In short: use Google Lens “About this image”, a C2PA verifier on the web, and a reverse search, then read the pixels themselves.
First, try the Lens route, which Google documents on the same help page. Open the photo in the Google Photos app, tap More, then Google Lens, swipe up, then tap “About this image”. That view shows where else the image appears online, which often settles a “seen this before” question faster than the panel does. A photo that turns up on a stock site or an old news page is not the fresh, personal shot it claimed to be.
Second, read the same signed record on the web. The panel is not available on Google Photos web, but the record itself is a standard, so other tools can read it. Adobe’s free Content Credentials verify tool at verify.contentauthenticity.org opens the file and shows the same history the panel would show. Upload the file itself, never a screenshot of it.
Third, reverse search the photo. A panel that says nothing and a reverse search that finds the photo on five other sites is a clear answer. A panel that says nothing and a reverse search that finds nothing is still an open question, and you should say so rather than calling the photo real.
Fourth, read the pixels. The panel says nothing about manual retouching, and it says nothing about AI edits made by tools that write no record. Hands, ears, text in the background and light that falls the wrong way are still worth checking by eye. Our guide to spotting a photoshopped photo covers the manual tells that no metadata check will ever catch.
Fifth, remember the other Google route. Google’s help page notes that SynthID watermarks embedded in images edited with generative AI can surface in Google Search and Lens. Our guide to telling whether an image is AI generated puts the panel, the SynthID route and the visual checks together in one place, so you can work through them in order instead of guessing which one to trust.
Common mistakes when reading the panel
In short: the panel is easy to misread, and most mistakes come from checking the wrong file or the wrong screen.
- Checking on the web. The panel does not exist on Google Photos web. If you check there and see nothing, you have learned nothing about the photo.
- Checking a screenshot of the photo. A screenshot is a new file with no history. Check the original file, not a picture of it.
- Editing the photo first, then checking it. Google Photos adds Content Credentials when an AI edit is saved, even if the original file had none. Check the untouched file first, or the panel will show the edit you just made rather than the history you wanted to test.
- Reading “May have been edited by AI tools” as proof. It is a maybe. It means an app recorded an edit without saying whether AI was used.
- Treating a clean panel as proof the photo is real. The Opera House photo in the worked example is real and shows no panel at all. A fake photo with no history looks exactly the same.
- Expecting the label to follow the photo to Instagram. Instagram reads the file’s metadata when you upload and may add its own label, but that is Instagram’s reading, not Google’s panel. The Instagram AI info label guide explains what trips that label and how to check your file before you post. Our guide to AI labels on social media explains which edits trip the label and which do not.
FAQs
Is the โHow this was madeโ panel available on Google Photos web?
No. Googleโs help page states that Content Credentials are not available on Google Photos web. The panel only appears in the Google Photos app on Android and iOS. If you only have a browser, upload the file to a C2PA verifier instead, and check the original file rather than a screenshot of it.
Why does my photo not show โHow this was madeโ?
Most photos carry no Content Credentials, so the section simply does not appear. The photo may predate the standard, come from a camera or app that never writes it, or have passed through a chat app that stripped the metadata. An empty panel is the normal state, and it proves nothing about the photo.
Does โEdited with AI toolsโ mean the whole photo is fake?
No. The line records that generative AI created new pixels somewhere in the photo, for example when Magic Eraser removed an object. One small edit triggers the same line as a large rebuild. Read it as a record of tool use, then look at the photo itself to judge how much changed.
Can someone remove the โHow this was madeโ information?
The signed C2PA record is hard to alter quietly, because each editor signs its entry and tampering shows up as an error. But the record can be lost honestly: re-saving, screenshotting or sending the photo through a chat app can strip it. IPTC โAI infoโ metadata is easier still to edit or remove.
Does a photo edited with Magic Eraser always show the panel?
Only when the edited file keeps its Content Credentials and you open it in the Google Photos app. Google Photos adds credentials when an AI edit is saved, even if the original had none. But if the file is then screenshotted or passed through an app that strips metadata, the panel will be empty.
My photo says โMay have been edited by AI toolsโ. What should I do?
Treat it as a maybe, not a verdict. A third-party app recorded an edit without saying whether AI was used. Check the same file with a C2PA verifier, run a reverse image search to see where else it appears, and look for visual tells such as warped hands or mismatched light.







