Key takeaways
- AI raises the floor of creator work and does not move the ceiling. The result is not a flood at the top, it is the collapse of the commodity middle.
- Three of the four stages of creator work got much cheaper. Judgement, the fourth, did not move at all.
- Competent, generic, unopinionated content used to be a business. It is now free, which makes it worthless rather than merely cheap.
- Answer engines do not punish AI content for being AI. They ignore it for being undifferentiated. Lived specifics are the moat.
- Automate production. Never automate what you believe, the example from your own experience, or the decision about what not to say.
Quick answer
Both. AI raises the floor of creator work dramatically and does not move the ceiling at all. The result is not a flood at the top and it is not free leverage for everyone. It is the collapse of the commodity middle, where competent, generic, unopinionated content used to earn a living.
The two loudest positions on this are both wrong. One says AI is a flood that will bury creators under infinite generated content. The other says AI is pure leverage and now one person can output like a team. The first ignores that nobody was struggling to find more content. The second ignores that output was never the bottleneck.
Here is what actually changes.
What did AI actually change about creator work?
Break creator work into four stages and the effect is uneven in a very specific way.
| Stage | Effect of AI |
|---|---|
| Research and synthesis | Enormously faster. What took a day takes 40 minutes. |
| Drafting | Faster, and quality-capped. It reaches competent quickly, then stops. |
| Production | Transformed. Editing, transcription, repurposing, subtitles, formatting. |
| Judgement | Unchanged. It never had the inputs to contribute here. |
Three of four stages got much cheaper. The fourth did not move. That asymmetry explains everything downstream.
Why does the middle collapse?
Content used to sort roughly into three tiers. Bad, competent, and distinctive.
The competent tier was a real business. Well-researched, clearly written, professionally produced, no particular point of view. Plenty of people made a living there, because competent was scarce enough to be worth paying for.
AI made competent free and instant. So the tier that used to pay is now the tier that anybody produces at zero cost, in volume, on demand. It has not become worse. It has become worthless, which is different and worse.
What survives is on either side of it. Genuinely distinctive work, which requires having done something and having an opinion about it. And genuinely useful utility, which requires being right rather than interesting. The middle is gone and it is not coming back.
The operator move: If a reader could get the same thing from a chatbot in ten seconds, do not spend your week on it. That single test reprices your entire content calendar.
Is AI-generated content punished by search and answer engines?
Not for being AI-generated. It is punished for being undifferentiated, and most AI content is undifferentiated, so the outcome looks the same from outside.
The thing worth understanding is that discovery itself changed shape. People increasingly ask a question and get an answer rather than a list of links. That shifts what gets rewarded.
- Direct answers beat build-up. If the answer to the question in your headline is not in the first 50 words, an answer engine has nothing to lift.
- Specific, checkable claims beat general ones. Numbers, ranges, named mechanisms. Vagueness is unquotable.
- First-hand experience beats aggregation. Every model can summarise the consensus. None of them can report what happened to you in a meeting last Thursday.
- Structure beats prose flow. Question headers, clean lists, definitions that stand alone. This is unglamorous and it is how you get cited.
The uncomfortable implication: the work most likely to be surfaced by an AI system is the work an AI system could not have produced. Lived specifics are the moat, and they are the one input that cannot be synthesised.
What should creators automate, and what should they never automate?
My working split, after running this across both team work and my own publishing.
| Automate without hesitation | Never automate |
|---|---|
| Transcription, subtitles, clip selection | What you believe and why |
| Repurposing one piece into other formats | The specific example from your own experience |
| Research gathering and first-pass summarisation | The decision about what not to say |
| Formatting, metadata, alt text, scheduling | The final line, which is where voice lives |
| Headline and hook variants, as options not answers | Any claim you have not personally verified |
The test is simple. If a reader could get the same thing from a chatbot in ten seconds, do not spend your week on it. If they could not, that is the only work worth doing. I went deeper on the practical side of this in how to use AI for content creation without sounding like AI.
Does AI make it easier or harder to start as a creator?
Easier to start. Much harder to stand out. These are not in tension, they are the same fact viewed from two ends.
Production quality is no longer a barrier. A person with no budget can now publish work that looks and sounds professional from day one. That is a real democratisation and it is worth celebrating.
But when production stops being a filter, everything rests on the thing production used to partly disguise: whether you have something to say. More people can now clear the technical bar, so clearing it distinguishes you from nobody.
Practically, this means the sequence has inverted. It used to be: learn to produce, then find your voice. Now it is: find something you actually know, then let the tools handle everything else.
What happens to the creator economy from here?
Three things I would bet on.
Audience size matters less, audience specificity matters more. When content is infinite, the scarce thing is being trusted by a particular group about a particular subject.
Proof of having done the thing becomes the differentiator. The operator who publishes will beat the commentator who publishes, by a widening margin, because the operator has inputs that cannot be generated.
Volume strategies stop working. Publishing more was a viable edge when producing was expensive. It is not an edge when it is free for everyone. The next edge is being right, being specific, and being someone.
The operator move: Look at your last five pieces. How many could a model have produced from public information in under a minute? That number is the share of your work that stopped being worth doing.
Frequently asked questions
Will AI replace content creators?
It replaces the production work, not the judgement. Creators whose value was competent execution are genuinely at risk. Creators whose value is a point of view built on doing the work are not.
Should creators disclose that they use AI?
Disclose it for anything where a reader would feel misled to learn it, particularly synthetic voice, video, or fabricated first-person experience. Nobody needs a disclosure that you used a tool to transcribe an interview.
Does AI content rank in Google?
Content is not penalised for how it was made. It is penalised for adding nothing. Most AI content adds nothing, which is why the correlation looks like causation.
What is GEO and why does it matter for creators?
Generative engine optimisation is structuring content so AI answer systems can find, quote, and attribute it. Practically it means direct answers near the top, specific checkable claims, clear question headers, and first-hand detail that cannot be synthesised.
Is there still money in the creator economy if AI floods it?
Yes, and it concentrates. Generic gets cheaper, specific gets more valuable. The people earning well in three years will have narrower audiences and stronger opinions than the people earning well today.
Leave a Reply