Answer
How do you stop AI making your work generic?
Supply the particulars it cannot know: your examples, your constraints, your position. Generic output is what remains when nothing specific was given.
Supply what makes your case particular rather than describing it. Generic output is the average of everything written on a subject, and it is what remains when nothing specific to your situation was given to work from.
The output is generic for a structural reason. Given a subject and no particulars, the most probable response is the median of everything written about that subject, which reads as competent, balanced and empty. That is not a failure of instruction and it will not be corrected by asking for originality, because the missing thing is material rather than intent.
The first correction is your own examples. A previous piece you were happy with carries your structure, your emphasis, your habitual moves and your register, and supplying it transfers all of them at once. This is the difference between output that sounds like your business and output that sounds like the category, and no description of a voice achieves it. Three examples do more than any amount of adjectives.
The second is the particular facts. What you actually charge, what you actually did last quarter, the specific customer who had this problem, the exact constraint you are working under. Generic writing is generic largely because it contains no particulars, and supplying them changes the output more than any instruction about tone. This is also what makes writing worth reading, which is a happy alignment.
The third is your position. Most subjects have a conventional treatment and yours may differ, and the difference is the reason anyone would read yours rather than the conventional one. Stating what you think — including where it departs from the usual advice — gives the output something to be built around. Without it, the conventional treatment is the only available shape.
The fourth is removing the padding that generic writing conventionally includes. The scene-setting introduction, the balanced acknowledgement of the other side, the summary of what was just said, the closing exhortation. These arrive by default because they are what the form contains, and each one dilutes the specific content around it. Cutting them is usually the single largest improvement to a draft.
One structural point about repeated use. A business producing regular content this way accumulates pieces that resemble each other, because each was produced from a similar brief by a similar process. The individual piece can be fine and the corpus becomes uniform, which is invisible when each is reviewed alone. Comparing a new piece against what you have already published is the only stage at which that is catchable.
Generic is not a style the system chose; it is the shape of an answer with nothing particular in it.
Siddharth Sharma, Context Theory
Related questions
Does asking for a specific tone help?
Less than showing one. A described tone produces an approximation of a category — professional, conversational, authoritative — and those categories are themselves generic. A sample of your own writing conveys something no category name can, because your register is not one of the available labels.
Is it worth using AI for work where voice matters?
For drafting and structuring, yes, provided the particulars go in and the voice comes from your own examples. For anything where the value is the perspective rather than the content, the useful contribution is usually the argument you are missing or the objection you have not considered, rather than the prose.
METHOD
Every figure below carries its source and the date it was verified. Nothing on this page is asserted.
The numbers on this page.
| What | Value | Specific to |
|---|---|---|
| Ranking loss for scaled near-identical page farms | 60–90% | Category-wide |
| AI-cited sources that also rank in the Google organic top 10 | 10% | Category-wide |
Google March 2026 core update — scaled content abuse · verified
2026 generative engine citation study · fewer than · verified
What is specific to this page.
| Kind | Claim | Check it against |
|---|---|---|
| Workflow | Given a subject and no particulars, the most probable output is the median treatment of that subject, so generic writing is the structural result of an absent input rather than a stylistic choice to be instructed away. | Comparing output on a subject with no supplied particulars against the same request with specific facts included. |
| Response | A sample of the writer's own previous work transfers structure, emphasis, habitual moves and register simultaneously, which no description of a voice achieves because the register is not one of the available category labels. | Comparing output produced from a described tone against output produced from three samples of the writer's own work. |
| Software | Conventional padding — scene-setting openings, balanced acknowledgements, restatements and closing exhortations — arrives by default because it is what the form contains, and each element dilutes the specific content around it. | Removing every paragraph that adds no fact or argument from a draft and comparing what remains. |
| Constraint | Repeated production from similar briefs by a similar process yields a corpus of resembling pieces even where each piece is individually acceptable, and this is invisible when pieces are reviewed in isolation. | Comparing a new piece against the business's existing published work on adjacent subjects. |
Each row would be wrong on another industry's page. Where a sourced figure exists it is in the table above instead; these are the constraints that shape the work and do not happen to be numbers.
Start with the measurement.
Reading about a benchmark is not the same as knowing your own number. The audit produces yours, measured rather than estimated.
$497 · delivered in 5 business days · credited against month one