Key Findings
- Proving an image is real is now a part of some photographers’ jobs.
- People distinguish real photographs from AI-generated ones at about the rate of a coin flip.
- Provenance labels on news images improves viewer's trust in them.
- Photographers want AI disclosure but reject a blanket "AI used" flag.
- Photographers have among the highest career optimism of all creative fields.
Cheriss May was one of three photographers on assignment for Getty inside the Capitol on January 6, 2021..The following day, back at the Capitol to retrieve her bag, she encountered a man outside the gates who told her the events of the previous day had not happened. He called it a hoax.
“I told him I was there yesterday and I can show you the pictures,” she said in a recent interview with Adobe Labs. “It wasn't a hoax. I saw it with my own eyes. Not only that, but my second set of eyes saw it too.”
The second set of eyes is her camera, and she means the phrase precisely. “I view the camera as my second set of eyes. It's also very genuinely pure and real because there's no editorial. You can't editorialize the camera. It records what is in front of it.”
May is a Washington-based portrait and editorial photojournalist and a professor of visual journalism at Howard University. She has photographed politics and protests for years. Her work has run in The New York Times and People. She is a member of the White House press pool.
That exchange changed how she works. She is working in a profession where the foundation has shifted beneath her.
“The question used to be, did you get the shot,” she says. Now the question is, “Is the photo real?”
“I fell in love with storytelling”
May was seven or eight when her mother, a business teacher who advised the high school yearbook, brought a camera home. May asked to play with it.
“Photography really found me,” she said. “It started with a childhood curiosity and then it was a hobby, then it was a serious hobby, and then it eventually became my profession.”
She became her family’s photographer, often with the camera pressed to her face. What she remembers is not taking the pictures but what came after: collecting the prints, laying a cloth over the table, and setting the pictures out for everyone to look through. “My family would look through the photos and I was seeing the joy that it brought them. Taking these photos created this love for photography for me.”
That was decades before she started to cover the White House. May started out as a graphic designer for several newspapers, including The Washington Post and Stars & Stripes. She later worked for USA Today’s sports desk. After she was laid off, she started to consider how else she could apply her visual passion. Photography had always been running alongside the job, so she decided to find out what it could become.
What she found was not exactly photography, but a method for capturing meaningful moments. “When I started covering politics, when I started going to the White House, I fell in love with storytelling. It was less about taking pictures. I feel a responsibility to document these historical moments, and to give people a window into these places, behind the curtain of what you normally don't see.”
Images of the president don’t always do this. One of her favorite examples came from when she photographed a voting-rights campaign for the NAACP. She was in Washington and had set up a small mobile studio in a meeting hall with a backdrop and lights.
During a break, a woman who cleaned the building came over and asked what was going on.
After May explained what she was photographing, the woman shared that she never liked how she looked in pictures. May told her to come back when she finished her shift. When she did, May positioned her against the backdrop and shot tethered, so the frames appeared on her laptop.
“I invited her over to come look, and she just started crying.” May asked why. “She said, ‘I've never seen myself like that.’ For me, it was being able to show someone just how powerful they are, just how extraordinary they are.”
“Professional photographers have the highest career optimism of any creative role, according to a study of 1,952 creatives in three countries.”
This pride in the craft is common among photographers. In a study of 1,952 creatives across the US, UK, and Japan, photo professionals reported the highest career optimism of any creative role, at 87%.
Even with changes in creative fields, May expects the world to need photographers long into the future. She is not concerned about being replaced; she is more concerned about being believed.
How May builds credibility into her images
May used to have a reliable trick for spotting a generated image.
“If somebody came to me and said, ‘Hey, is this photo real?’ the first thing I would do is look for the hands.” Hands are complicated and detailed and were long thought difficult to fabricate. But when a model is trained on thousands of hands, the result is considerably more realistic. “I can't say that I can easily look at something and say that's fake.”
She is more candid about that than many in her industry. In interviews with 20 photo editors and visual journalists across 16 news organizations in seven countries, researchers at RMIT, Washington State, and Queensland University of Technology found that many believed they could spot AI-generated images better than the general public could. The evidence on viewers suggests nobody should be confident. A September 2025 survey found people correctly identified AI-generated images 52% of the time and real images 49%, roughly a coin flip. In a larger study, more than 12,500 participants judging approximately 287,000 images were only marginally better, at 65% for faces and 59% for landscapes.
Since spotting subtle clues is no longer a reliable method, May's answer is to build a record that doesn't depend on anyone's eye, including her own.
That record has a standard behind it. C2PA is the open technical standard behind Content Credentials, published in 2021 and supported across camera manufacturers, platforms, and editing tools. Adobe co-founded the Content Authenticity Initiative in 2019 and helped establish the standard. It produces a record of provenance, a verifiable account of where an image came from and how it has been modified, attached to the file and checkable by whoever receives it.
The underlying idea of C2PA is spreading in different formats. Earlier this month, Apple introduced its own photo authenticity feature that writes a signed identifier into a photo's metadata as evidence that a camera sensor, and not a model, produced it.
How You Can Prove Your Photography Is Your Own Work
You don't need a newsroom's resources to build evidence that your work is your own. Content credentials based on the C2PA standard can be attached to a digital image to demonstrate where it came from. Three steps are critical:
1. Capture with credentials switched on. This ensures that authenticity is validated at the moment of capture. The most reliable way is to capture images on a camera or phone that natively supports the C2PA standard in hardware. If your phone does not contain C2PA support in hardware, you can use a C2PA-enabled capture app that signs the photos in software instead. This is not as indelible as hardware, but it still starts the record.
2. Record your edits. Edit in software that writes your changes into the credential, so adjustments like exposure, color, and cropping become part of the record. Editing platforms that support Content Credentials include Adobe Lightroom, Photoshop, and Capture One.
3. Attach your name, then publish. Before you share the image, use a Content Authenticity tool to attach your identity to it. Link your name and accounts, such as a personal site, LinkedIn, or Instagram, so viewers can confirm who made it. You can also add a request that the image will not be used to train AI. Then export or upload the file with the credential attached, so anyone who receives it can open the record and check it for themselves. The same tool lets you add credentials to older images you shot before Content Credentials existed.
May follows a content authenticity process much like the one shown above and adds additional safeguards of her own. She keeps the untouched original (RAW image) of every frame she captures. She intentionally captures images around a moment as well as the moment itself, including frames before and after and a wider shot of the surroundings. As she works, she also records a second video stream alongside her stills, giving her, as she puts it, “an even deeper chain of custody.”
The last part of her method is older and plainer than any tool: being ready to show the work. “It's going to take time to build the trust that you are someone that is transparent…someone who is not afraid to share that information” Based on her experience, people are increasingly asking for proof.
Under the hood: how content credentials are made and checked
Although adding content credentials when capturing a digital photograph is crucial for proving its provenance, you can add content credentials at any stage of an image's lifecycle. The process for making and checking content credentials is consistent in all use cases.
Stage
What does it
What happens
1. Sign the record
Your camera, phone, or editing software.
Gathers the details of how the image was made and seals them together with a tamper-evident digital signature, backed by a trusted authority.
2. Attach it to the photo
The same camera, phone, or editing software.
Tucks that sealed record into the image file (or keeps it in a small companion file) and ties it to the picture itself, so any later change to the image shows up.
3. Check it
Any app or viewer that reads Content Credentials
Opens the record, confirms the signature is genuine and unbroken, checks that it still matches the image, and shows whether it passes.
“Don't even trust me”
May does not ask people to trust her. She asks for the opposite.
“I'm not saying trust me as the photographer. What I'm saying is trust your own eyes. Look at the provenance of this image I captured and see for yourself. This photo hasn't been doctored…you can see that for yourself and make the determination.”
There is now evidence that this system of provenance works on viewers. In a peer-reviewed experiment with 6,114 participants across the US, UK, and Norway, researchers at MediaFutures and the University of Bergen added C2PA provenance labels to news images and found that readers judged those images more credible and reported greater trust in the source. Most arguments about how to handle AI-generated media rest on reasoning about what ought to work. This one rests on evidence that it does.
How much the record says matters. May has been clear that a broad “AI used” label would work against her. A generic flag, applied without distinction, reads as manipulation regardless of what was actually done. “If you say AI, no matter what the usage is, then people are not going to trust that what they're seeing hasn't been manipulated,” she told Adobe in an earlier conversation.
She is not alone. In a study with 34 professional and semi-professional photographers in the US and Europe, the photographers sorted 26 editing tools into “always show,” “let me decide,” and “never show.” They supported disclosure in principle and objected to a single undifferentiated label covering everything from noise reduction to background replacement. Photojournalists leaned furthest toward full transparency, describing a credential that names specific actions as one of the most powerful authenticity tools.
As May and many of her colleagues see it, disclosure that says what was actually done is an asset to a working photographer. Disclosure that flattens every action into a single flat label is not.
What widespread image credibility will take
May does not think this is a technology problem with a technology answer.
“Technology helps, but it's going to take the mindset of people being open to wanting to verify things for themselves,” she says. “And the more transparent we are, the more that is going to help people understand the importance of that.”
Part of the burden is on photographers, she says. Part of it should sit with platforms too. Platforms can standardize the provenance process and make information visible and universally understandable.
“I would love to see a world where you can immediately tell by some kind of color label on the post or the story, and then you put your cursor on it, and it gives you that nutrition label of the provenance of that image. I want it to be a part of the fabric of our lives, so that it's not something you have to search for.”
C2PA also uses the term nutrition label to disclose a credential. It’s not a verdict or judgment, but a quickly scannable set of information that anyone can see to judge for themselves.
Adoption of C2PA is concentrated in professional photojournalism infrastructure, in cameras and newsroom workflows, rather than across the wider internet. Provenance does not retroactively authenticate an anonymous post, and the absence of a credential is not evidence that an image is fake, since most images carry no record at all. What a credential does is let someone who did the work prove it.
That picture is starting to change beyond the newsroom. LinkedIn, TikTok, Instagram, and Google Search can now surface the credentials associated with an image to everyday viewers, and a growing set of consumer cameras and phones can create one. As that capability spreads, anyone will be able to show that a photo is theirs, and viewers will be able to review the built-in record.
Adobe researchers are also developing features like invisible watermarking and image fingerprinting to maintain an image's provenance record, even after it's been resized, screenshotted, or reposted. Learn more about that work here.
For May that is enough for now, because it means no one has to take her word for the authenticity of her work.
Known limitations
This piece profiles one photographer. Her workflow is her own and is not presented as typical of the profession.
The Adobe research cited here is qualitative and not representative of population-size trends. The 34 participants plus four interviews in the June 2026 study do not support claims about photographers as a whole, and the study's segment-level findings do not rest on samples large enough to characterize those fields. It covered the US and Europe only.
The findings on creative professionals' attitudes come from a study of approximately 1,952 creatives conducted in the US, the UK, and Japan. The figures cited for photo professionals describe a subgroup of that sample rather than the full set, and should be read as directional. All of its measures are self-reported. Its geography also differs from the June 2026 study, so the two should not be read as describing the same population.
The newsroom interviews cited here were conducted between March and November 2023, before the current generation of image models, and they record what photo editors believed about their own ability to spot AI images rather than a test of that ability.
The external studies on how well people distinguish AI images use different samples, image sets, and methods, and their percentages should be read as a consistent direction rather than as directly comparable figures. The provenance-label experiment measured judgments of news images in a controlled setting, which is not the same as behavior in an ordinary feed.
Methodology
This article is built on an interview with Cheriss May conducted on August 11, 2026, supported by an internal Adobe research study, a study conducted with Bond Brand Loyalty, and several external sources. Quotes have been lightly edited for length and for spoken repetition. The meaning has not been changed.
The June 2026 study was a two-part, mixed-method study conducted by Adobe Design Research & Strategy. Part one was a 15-minute unmoderated card sort with 34 professional and semi-professional photographers in the US and Europe, fielded May 26 to 30, 2026. Part two was four 30- to 45-minute interviews with full-time photographers spanning photojournalism, commercial, fine art, and news photo editing, fielded June 1 to 4, 2026. Cheriss May was one of the interview participants.
The newsroom findings come from a peer-reviewed study published in Digital Journalism, in which researchers at RMIT University, Washington State University, and Queensland University of Technology interviewed 20 photo editors and staff in equivalent roles across 16 news organizations in Australia, France, Germany, Norway, Switzerland, the United Kingdom, and the United States. Interviews ran 30 to 90 minutes and were conducted between March and November 2023.
Sources
Interview with Cheriss May, portrait and editorial photojournalist, conducted by Paul Slater in Washington, DC on August 11, 2026.
Adobe research on creative professionals' attitudes toward AI, ownership, and consent, including 1,952 respondents in the US, UK, and Japan. 2026.
Thomson, T. J., Ryan J. Thomas, and Phoebe Matich. Generative Visual AI in News Organizations: Challenges, Opportunities, Perceptions, and Policies. RMIT University, Washington State University, and Queensland University of Technology. Digital Journalism, vol. 13, no. 10. 2024.
Can consumers distinguish between real and AI images in 2025? Conjointly. September 25, 2025.
Roca, Thomas, Anthony Cintron Roman, Jehú Torres Vega, Marcelo Duarte, Pengce Wang, Kevin White, Amit Misra, and Juan Lavista Ferres. How Good Are Humans at Detecting AI-Generated Images? Learnings from an Experiment. Microsoft. arXiv. May 2025.
Coalition for Content Provenance and Authenticity (C2PA). Open technical standard. 2021.
Content Authenticity Initiative (Adobe and founding partners). 2019.
Trattner, Christoph, Svenja Lys Forstner, Alain D. Starke, and Erik Knudsen. C2PA Provenance Labels Increase Trust in Digital News Platforms Across Western Countries. MediaFutures, University of Bergen. Proceedings of the International AAAI Conference on Web and Social Media, vol. 20, no. 1. 2026.
Adobe research on how professional photographers want AI-assisted and generative editing disclosed in Content Credentials, comprising a 34-participant card sort and four follow-up interviews, Lirra Hill, Staff Experience Researcher, Adobe Design Research & Strategy; June 2026.
Collomosse, J. and Agarwal, S. "This World Photography Day, Adobe Research Maps How Trust Travels with a Photo." Adobe Research. August 19, 2026.
Acknowledgments
This article was produced by Adobe Labs. Paul Slater conducted the interview with Cheriss May. The June 2026 photographer research was led by Lirra Hill, staff experience researcher, Adobe Design Research & Strategy, with contributions from Noga Hurwitz, Coleen Jose, and Andy LeMay. Meredith Kunz, Pei Ketron, and Josephine Lee provided content and technical review. We thank Cheriss May for her time and candor discussing her work.
