Here is a sentence that should not need to be radical:
If a community trusted me enough to show up in the room, I am not going to replace them later with synthetic almost-people because the recap needs a stronger hero image.
I can make a beautiful fake person in fourteen seconds.
Perfect skin. Dramatic light. Inclusive outfit math. The whole thing glowing with the warm dead-eyed confidence of a brand deck that has never had to ask anyone how to spell their name.
Still not a portrait.
A portrait is not an image of a face.
A portrait is a record of attention between people.
It says: you stood there, I stood here, something happened between us, and this is the trace.
That trace matters more now, not less.
In the synthetic age, the fact that somebody actually showed up becomes part of the value of the work.
So here is one of my lines:
I will use AI around documentary work.
I will not use AI to fake the people who showed up.
Documentary Work Is Not A Pixel Problem
The dumbest version of the AI conversation treats every image as output.
Image goes in. Prompt goes brrr. Image comes out. Save money. Ship faster. Congratulations, you have invented visual karaoke with a procurement department.
That worldview only makes sense if you think photography is a production step.
I do not.
I come from documentary photography, community media, event documentation, artist portraits, festival chaos, hallway conversations, and all the weird human moments that do not fit inside a shot list.
The job is not “make picture.”
The job is witness.
Sometimes the job is to make someone feel safe enough to be seen.
Sometimes it is to notice who keeps getting left out of the frame.
Sometimes it is to catch the moment where a room stops performing and starts telling the truth.
Sometimes it is to build an archive so a community can remember itself later, with names attached, relationships intact, and enough context that the future does not have to guess what mattered.
AI can generate a plausible image of community.
It cannot be accountable to that community.
It cannot remember who was nervous before the talk. It cannot ask the elder how they want their name spelled. It cannot feel the room shift when ceremony changes the order of operations. It cannot know that the person in the back corner built half the thing but will never walk toward the microphone unless someone notices.
That is not nostalgia.
That is the work.
What AI Can Do Around The Frame
None of this means “keep the machines out.”
Please. I am not trying to turn documentary practice into a candlelit museum of analog purity where everyone has to churn their own film chemistry and speak only in contact sheets.
AI can be wildly useful around the frame.
It can help me search thirty years of images when my memory has the feeling but not the filename.
It can help draft captions from notes.
It can turn a transcript into a clip list.
It can pull recurring themes from interviews.
It can help build contact sheets for a client.
It can organize names, projects, roles, places, and relationships so the archive becomes a living map instead of a hard drive full of ghosts.
It can help generate a shot list before an event, identify missing proof after an event, and turn the whole thing into a useful story package that organizers, funders, participants, and future collaborators can actually understand.
That is fair game when the consent, privacy, and context are right.
That is the 75% Rule in documentary practice: send the machine after the art-adjacent work.
Let it carry boxes.
Let it sort cables.
Let it find the quote.
Let it rename the folders.
Let it help the human witness travel farther.
But do not confuse the logistics around witness with the witness itself.
The Line I Draw
Here is the line in plain language.
I will not replace a portrait of a community member with an AI-generated portrait of a community member.
I will not create fake documentary evidence of a gathering.
I will not invent attendance, diversity, intimacy, consensus, ceremony, joy, grief, or solidarity because a deck needs a better slide.
I will not let “close enough” become the standard for human truth.
That last one is where the danger lives.
Synthetic media trains organizations to accept plausible surfaces.
Looks like a person.
Looks like a crowd.
Looks like a community.
Looks like care.
But documentary work is not about looking like care.
It is care with a shutter attached.
When an organization uses synthetic people to represent a real community, it is not just saving money. It is practicing the idea that people are interchangeable symbols. It is rehearsing extraction in a warm color palette.
That is how trust leaks out of a culture.
Quietly.
Pixel by pixel.
Presence Is Provenance
AI people love the word provenance right now.
Good. We need it.
Source data. Model lineage. Watermarks. Chain of custody. Labels. Receipts. Bring me the receipts. I am pro-receipt like a retired accountant with a grievance.
But documentary practice has always carried another kind of provenance:
Were you there?
Did they know why you were there?
Did you ask?
Did you listen?
Did you spell the name right?
Did you send the photo back?
Did the people represented get any value from the representation?
Did the archive become a commons, a trophy case, a liability, or a gift?
In the next few years, that kind of provenance is going to matter more than ever.
Not because every photograph needs to become a legal deposition.
Because the internet is about to be flooded with images that never touched a life.
When everything can be generated, the lived trace becomes precious.
The proof that someone stood somewhere, on a territory, inside a particular weather system of relationships, will become part of the meaning.
Presence becomes provenance.
Questions Before You Replace Documentation
If you are an event producer, funder, nonprofit, school, studio, public agency, arts org, or community group thinking about AI-generated imagery instead of real documentation, slow down for ten minutes and ask better questions.
What job is this image doing?
Is it illustrative, conceptual, speculative, or documentary?
Will anyone reasonably read it as proof that something happened?
Are we representing real people, real communities, real cultural practices, or real relationships?
Who benefits if we fake it?
Who disappears if we do?
What would happen if we hired someone from the community to document it instead?
Could AI help us prepare, organize, caption, archive, and distribute the real documentation rather than replace it?
If the answer is “we just need a vibe,” say that out loud.
Sometimes a vibe is fine. Use illustration. Use abstract imagery. Use clearly labeled speculative art. Use weird generated textures until the walls melt.
But if the image is standing in for real human presence, treat it like a promise.
Do not make promises with fake people.
Both Hands Full, Again
This is the same contradiction because apparently this era has decided contradiction is its native file format.
AI is built on a messy pile of consent problems, power problems, labor problems, bias problems, and cultural memory problems.
AI is also useful.
Both things are true.
So I am not walking away from the tools. I am also not handing them the sacred center of my practice and calling that adaptation.
I want AI in the archive.
I want AI in the notes.
I want AI helping me find the forgotten story, surface the buried pattern, draft the first pass, organize the proof, and make the real work easier to share.
I do not want AI replacing the person in front of me.
I do not want synthetic community to stand in for actual community.
I do not want documentary practice turned into “content operations” for institutions that want the glow of human connection without the responsibility of relationship.
The future is going to be full of generated everything.
Fine.
Then let the real thing get more real.
Let the rooms matter.
Let the people matter.
Let the photographer be more than a content machine.
And if someone trusted us enough to show up, do not fake them after the fact.