Getting AI to write content that ranks in Google and gets cited by ChatGPT is easy. Getting that content to sound like you is the hard part. You fix it by removing AI’s common tells, teaching it your voice with guidelines pulled from your own writing, and protecting your SEO structure while you do it.
We have been building content with AI at TJ Digital since ChatGPT launched, now across 40 to 50 client campaigns. That is how we deliver about four times the work at the same rates as a typical agency. For some clients we run over 100 writing guidelines to keep every piece sounding like them.
Why AI content ranks but still sounds dry
AI content ranks when its structure matches how people search, and that same structure is what tends to make it read dry. The recipe for ranking is simple. Your title and headings need to closely match what someone is searching for.
Immediately after each heading, answer the question that heading poses. Right after the answer, give supporting facts. Those can be data points, quotes, or tables.
There is a little more nuance, but that is the gist. The problem shows up when you tell AI to write this way. The output often comes back dry and academic, with no warmth or personality.
Machines are happy to read dry academic writing. Your customers want writing that feels human. That gap only matters if you expect humans to read the content and care what they think about your brand, which covers most businesses.
Why telling AI to add personality backfires
Asking a model to “add more warmth and personality” usually makes the writing worse. It does add personality. The personality is unlikable and obviously written by AI.
Telling it not to sound like AI does not work either. You cannot ask a model to stop being itself. You have to be specific about what you want gone.
Which AI tells to remove first
You get much better results by naming the exact tells you want the model to drop. Vague instructions produce vague fixes. Specific callouts produce clean writing.
Here are the tells worth removing before anything else:
- Contrastive framing like “it’s not X, it’s Y”
- Empty adverbs and hedges like “absolutely” or “actually”
- Trying to force every line to land
- Em dashes
Give the model enough of that direction and it will stop sounding like AI. It still will not sound like your brand yet. That takes a different step.
Why feeding AI your writing samples doesn’t work
Handing the model your old articles and asking it to copy them leads to repetition. It latches onto surface patterns and misses what you value in your own writing. It does not know whether you like the topics, the specific phrases, or the length of your sentences.
The result is a model that reuses the same phrases over and over. Everything it writes starts to sound the same. There is a better way to hand off your voice, and the comparison below shows why:
| Approach | What you give the AI | What you get back |
| Feeding writing samples | Old articles and posts as examples | Repeated phrases and surface mimicry |
| Extracting guidelines | A curated list of rules from your writing | A consistent voice without copied lines |
How to build a guideline list from your writing
Extract the guidelines in a separate chat, then write with only those guidelines. This keeps the model focused on your rules and stops it from parroting your sentences. It is one piece of a larger content machine that turns your expertise into scalable content.
Here is the process we use:
- In a separate chat, give the model many examples of your writing.
- Ask it to extrapolate a long list of guidelines from your style.
- Manually curate the list, keeping the rules you agree with.
- When you write content, give it the guidelines with no samples or example phrases.
That single change moves the output a long way toward your voice. It still will not be perfect. The last mile comes from feedback.
How feedback makes AI sound more like you
Every time the model writes for you, give it detailed feedback and turn your edits into new rules. If you rewrite a section yourself, feed the new version back and have the model extrapolate what you changed. Then add those changes to your guideline list.
Over time you build a long, curated set of rules that captures how you write. For some of our clients that list runs past 100 guidelines. Frontier models handle large rule sets well, so this scales further than people expect.
There is a ceiling worth respecting. In one benchmark even the best models followed only about 68 percent of instructions at a density of 500 rules. Put your most important rules first so the ones that matter most keep getting followed.
How to keep your SEO while adding voice
Build the SEO structure first, then layer your voice on top, and make sure no voice rule undermines that structure. This is the part people get wrong. They chase personality until the writing stops answering the search.
Keep the load-bearing pieces intact. Headings match the search, the answer comes right after each heading, and supporting facts follow the answer. Your voice lives in how you frame and connect those facts. The structure that gets you cited stays untouched.
Model choice matters here too. In our testing, the AI model we use follows brand voice and formatting rules more reliably than the alternatives, which makes long guideline lists worth building.
Common questions about AI writing and brand voice
Can you just tell ChatGPT to stop sounding like AI?
No. A blanket request does almost nothing. Naming the specific tells you want gone, one by one, is what moves the writing.
How many writing guidelines do you need?
As many as it takes to capture your voice. Some of our clients have over 100. Frontier models handle large lists well as long as you prioritize the rules that matter most.
Does making AI content sound human hurt your SEO?
Only when your voice rules fight the structure. Keep the SEO skeleton in place, add voice on top of it, and your content stays both readable and citable.
What is the fastest way to start sounding like yourself?
Record a few minutes of yourself talking, get the transcript, and use it as your source material. Your own words carry your voice better than any prompt you could write.
Build a content system that sounds like you
Your expertise is the one thing AI cannot fake, and the method above puts it into everything you publish. We build content systems that sound like the business owner and get recommended by AI. Talk to our team and we will map out how to do it for your business.