Instagram admits AI slop won. Here's how creatives prove they're real
AI-made content flooded feeds in 2025. Now Instagram's head, Adam Mosseri, says the hard truth out loud: it's going to be easier to verify what's real than to catch every fake.
Meta's current tags miss plenty of synthetic media and sometimes flag real photos after minor retouching. Meanwhile, models are getting better at mimicking the "less perfect" styles that used to feel human by default.
What this means
Platforms will try to label AI, but Mosseri expects detection to get worse as models improve. The proposed fix: fingerprint real media at capture, then preserve that chain of custody as it moves through your tools and platforms.
Camera makers and software vendors are already moving. The industry is rallying behind CAI/C2PA standards for tamper-evident metadata that ties media to its source. Instagram will need to reliably read that data if it wants to steer feeds toward authenticity.
"Make it uglier" isn't the point-proof is
Mosseri argues that glossy, "perfect" content is cheap now, so raw and imperfect will signal truth. You'll see more unproduced, unflattering, in-progress posts. Whether or not you buy the aesthetic argument, the direction is clear: you need visible proof you made the thing.
Polish isn't dead. It just can't stand alone. The new currency is process, provenance, and personality-things AI struggles to fake consistently.
Your playbook to prove you're real
- Show process, not just outcomes. Post sketches, layers, node graphs, screen recordings, time-lapses, and version history. Share screwups, reworks, and why you made decisions.
- Capture a chain of custody. Use cameras and apps that embed provenance (C2PA/CAI). Keep RAW files and working files. When you post, include a quick "receipts" carousel: a metadata screenshot, early comps, and short clips from your timeline.
- Preserve metadata on export. Don't strip Content Credentials. Standardize a "cred-safe" export preset and test which platforms keep it intact.
- Label AI use proactively. If you used AI for ideation, cleanup, or references, say so. Clear boundaries build trust faster than pretending everything is 100% hand-made.
- Lean into human texture. Behind-the-scenes photos, voice notes, whiteboard pics, messy desks, bad lighting. Imperfections are cheap to produce and strong as signals.
- Go live and invite friction. Live sessions, Q&A, shared docs, real-time editing. The latency, stumbles, and on-the-spot decisions are authenticity markers.
- Publish a provenance policy. Pin a short note: what tools you use, how you verify work, where to check credentials, and how to report fakes. Make it easy for clients and fans.
Practical setup (1-hour checklist)
- Enable Content Credentials in your creation tools where available.
- Adopt gear/apps that support C2PA provenance.
- Create two export presets: one "portfolio-polished," one "process-proof" with embedded metadata.
- Record your screen by default when working. Save short clips of key moments.
- Build a 3-slide "proof pack" template: metadata screenshot, early draft, timeline/layers.
- Write a 3-sentence caption format: what you made, how you made it, what was hard.
How this plays on Instagram
Expect feeds to favor "real-feeling" content over perfect edits. Final renders still matter for your portfolio, but they won't carry your account alone. Consistent behind-the-scenes posts, small experiments, and proof packs will do more to win attention and trust.
If Instagram eventually reads provenance metadata well, you'll be ahead. If it doesn't, your audience will still have direct proof they can understand and share.
If you're a photographer or designer
- Photographers: Shoot RAW + provenance when possible. Share contact sheets, lighting tests, outtakes, and RAW-to-final comparisons.
- Illustrators/3D: Share sketches, layers, node trees, UVs, sim settings, and viewport captures. Post the messy first pass next to the clean final.
- Motion/video: Share timelines, proxies, LUT tests, grade toggles, and audio scratch takes. Export short edits that reveal your hand.
What to stop doing
- Posting only polished finals with no trace of how they were made.
- Over-smoothing or filtering everything to death. It wipes the human signals.
- Hiding AI use. The audience will assume worse than the truth.
The bigger picture
AI isn't going away, and platforms won't save you. You'll win by being harder to fake: public process, verifiable metadata, and a voice people recognize instantly.
The work is the proof. Show it while you build it.
Want structured practice with AI workflows?
If you need practical training to integrate AI without losing your signature, explore role-based options here: AI courses by job. For designers looking for hands-on, design-focused lessons, see AI Design Courses.
Your membership also unlocks:
