Why AI writes a generic bio, and what to feed it instead

The gallery came back three weeks ago. Fifty images. The plan was to finally post consistently, update the About page, rewrite the LinkedIn headline that has been wrong since the pivot.

The gallery is still open in a browser tab. The caption box is still blank.

So you paste a photo into ChatGPT and ask it to write something. What comes back could belong to any coach, any consultant, any woman with a good photo and a service to sell. You delete it and close the laptop.

AI writes a generic bio because it was handed a task without a brand. The model has no idea who you are talking to, what you sound like, or what that photograph was taken to do. Give it your ideal client, your voice, and the job each image is doing, and the writing stops reading like everyone else’s.

That is the whole fix. The rest of this explains what those three inputs actually look like and how to build them once so you never start from a blank box again.

Why does AI-written copy sound like everyone else’s?

Because nearly everyone types the same prompt.

“Write an Instagram caption for this photo.” “Write my LinkedIn bio.” “Make this sound professional.” The model does exactly what was asked, using the average of everything it has ever read about coaches, consultants, and service businesses. Average is the product you ordered.

Research out of the University of Washington found measurable linguistic convergence in AI-written content: brands running similar tools drift toward the same sentences over time. The same reporting notes that human-written content pulls 5.44 times more traffic. Two facts, one cause. A bare prompt carries no information about what makes you different, so the model fills that space with what is common.

Adoption is not the problem. As of April 2026, 87% of US small businesses use AI for marketing, up from 26% in 2023, and 81% of the businesses using it apply it to content creation. Nearly everyone has the tool. Almost nobody has given it a brand.

What does AI need from you before it writes a single line?

Three inputs. Skip any one of them and the output goes flat.

An ideal client, named specifically. Not “women in transition.” Not “small business owners.” A real person with an age, an industry, a budget, a fear, and a line you have actually heard come out of her mouth. The difference between “I help women in transition” and “I work with Bay Area consultants in their forties whose website photos are from 2019 and who wince every time someone clicks the About page” is the difference between copy that sounds like a template and copy that sounds like it was written for one person.

A voice, written down. Not “professional and approachable,” which describes roughly every business on earth. Real voice documentation looks like a list: words you use constantly, words you would never say, sentence length you favor, whether you swear, whether you use questions, what your opening move is. A list of banned words does more for output quality than a paragraph of adjectives.

The job each photo is doing. This is the input almost nobody gives, and it is the one that turns a photo library into a content engine. A photo of you mid-sentence in a client meeting is not the same asset as a photo of you at your desk with a laptop. One is proof that you do the work. The other is a background for a text overlay. When the model knows which is which, the caption writes itself around the image instead of floating next to it.

How do you use your photos as the input, not the decoration?

Start by giving every image a job before you write anything.

Go through the gallery once with a notes file open and label each image with what it proves and where it goes. Something like:

• Working with a client, mid-conversation, hands moving. Proves the work. Website services page, LinkedIn carousel about process.
• Standing, arms crossed, clean background, direct eye contact. Authority shot. Speaker one-sheet, press kit, LinkedIn profile photo.
• Laughing, seated, mid-gesture. Warmth. Instagram, email header, About page.
• Detail shot of hands, notebook, workspace. Texture. Filler slide, blog header, carousel background.

That labeling pass is one sitting and it does two things. It tells you what you have, and it gives the model something concrete to write from. “Write a caption for this photo” produces mush. “Write a caption for the image where I am mid-conversation with a client, for an audience of Bay Area founders who avoid posting because they hate how they look on camera, in my voice, opening with a line I have heard clients say” produces something you can edit rather than delete.

The photos are worth this effort. Buffer’s 2026 analysis of more than 52 million posts found images pull 2.3 times the engagement of text-only updates, and on LinkedIn a single-image post hits a 2.13% median engagement rate against 1.54% for text alone. The images are already doing the heavy lifting. The words just need to stop being an afterthought.

Can AI write your bio if you have never written one?

Yes, and this is the best use of it, because a bio is the piece most founders cannot write about themselves.

The trick is interviewing yourself first and letting the model do the shaping. Answer these out loud, record it, paste the transcript in:

• What did you do before this, and what made you leave?
• Who do you work with, and what do they usually get wrong before they find you?
• What is the thing you say to clients that nobody else in your industry says?
• What are you known for among people who have worked with you?
• What do you want a stranger to believe about you within ten seconds?

Then ask for three versions at three lengths: a fifty-word version for a speaker page, a hundred-and-fifty-word version for the About page, and a one-liner for the LinkedIn headline. Ask each one to be built only from the transcript, with no added adjectives. That last instruction is what kills the “passionate about helping others” filler.

You will still edit. But you will be editing your own sentences, rearranged, rather than staring at a blank page.

What about captions and emails?

Same rule, tighter loop.

For captions, the model needs the photo’s job, the audience, the voice file, and one constraint about form. Length caps work well. So does forbidding hashtags in the body, or requiring the first line to run under eight words so it survives the “more” cut.

For emails, the model needs one more thing: the actual outcome you want from the send. Booking a call is a different email from filling a workshop, which is a different email from bringing a cold list back to life. Name it in the prompt and the structure changes. Leave it out and you get a newsletter that ends in “let me know if you have questions,” which nobody answers.

Two habits raise the floor on everything:

1. Give it three of your own past pieces as examples. Real ones, even mediocre ones. Pattern-matching against your writing beats any adjective you could type.
2. Ask it what it is missing before it writes. A model that asks two clarifying questions produces better copy than one that guesses, every time.

Where does the time actually go?

Most small business owners are not short on effort. They are short on inputs.

Roughly 60% of small and midsize businesses spend one to ten hours a week on marketing, and that climbs to 72% for businesses with ten or fewer employees. That is a real chunk of a week, and a large share of it goes to the same repeated act: opening the caption box, writing four words, deleting them, and deciding to post tomorrow.

The block is almost never writing ability. It is the missing decision underneath. You cannot write a caption when you have not decided who is reading it. You cannot pick a photo when you have not decided what it is for. The blank box is a strategy problem wearing a copywriting costume.

What this looks like when the strategy already exists

Every client who comes through Gaby Clark Photography leaves with the strategy layer built, because it gets built before the camera comes out.

That is the ALIGN phase of the Align. Shine. Scale. Method™: an in-person strategy session where we define the ideal client and the brand voice together, in the room, on paper. No questionnaire sent home. No homework nobody fills out. By the end of that session there is a Personal Brand Blueprint and an Ideal Client Blueprint, which happen to be exactly the two documents any AI tool needs to stop writing like a stranger.

Then the SCALE phase hands over the tools that use them: a Canva template suite for LinkedIn headers, Instagram posts, carousels, covers, email signatures, and webinar slides, plus an AI Marketing Suite of eight named experts. Leo writes bios. Sonia plans social media. Carter repurposes content. Marta runs email campaigns. Megan writes website copy. Petra builds PR and media kits. Luca writes VSL scripts. Kane architects keynotes. Each one already knows the ideal client and the voice, because they were built from the blueprints made in the strategy session.

One client, a health coach, called wanting two headshots for a new website. On the phone we walked through what brand photography actually is: photos with clients, working out, eating well, the full arc of the work. By the strategy meeting, the uncomfortable part had surfaced. Years of thinking hard about what the work was, and not once about who it was for. We built the ideal client profile together. Every photo from that shoot had a job, and the marketing tools meant the images got used instead of archived.

That is the sequence. Clarity, then photos, then the writing that uses both.

Frequently asked questions

Can AI write a bio that sounds like me?

Yes, if you give it your own words to work from. Record yourself answering five questions about your background, your clients, and what you say that nobody else in your field says, then paste the transcript in and ask for a bio built only from that material with no added adjectives. The model edits your voice instead of inventing one.

Why does AI-written content sound generic?

Because a short prompt gives the model nothing to be specific about, so it defaults to the average of everything it has read. University of Washington research found AI content from different brands converges linguistically over time when the tools and prompts are similar. Specific inputs, meaning a named ideal client, a written voice guide, and a defined purpose, are what break the pattern.

Do I need professional photos to use AI for my marketing?

You need images with a purpose, and professional brand photos give you a library of them at once. AI can write around any picture, but the writing gets stronger when the image proves something: you doing the work, meeting a client, in the space where the business lives. A folder of headshot variations gives the model far less to say.

How long does it take to set this up?

Building the ideal client profile and the voice guide takes one focused session, and labeling a gallery is one sitting. After that, captions and emails run in a few minutes each. The setup is the cost. Everything after it gets faster, which is the opposite of how most marketing advice works.

Will people know my copy was written with AI?

They will if you skip the inputs and publish the first draft. They generally will not if the model is working from your transcript, your ideal client, and your voice guide, and you edit before posting. The tell is not the tool. The tell is copy that could belong to anyone.

The blank box is not a writing problem

The caption box stays empty because the decisions underneath it were never made. Once the ideal client is named and the voice is written down and every photo has a job, AI turns into something worth using: a fast, tireless drafter that already knows the brand.

Book a call at gabyclark.com and we will start where it actually starts, with who you are talking to.