Here is a quick test. Take the headline from your homepage, remove your logo and your company name, and paste it into a competitor’s site. Does anyone notice?
If the answer is no, you have met the real cost of AI generated content: not errors, not “robotic” text, but sameness. The copy is fluent, tidy and on-brief, and it could belong to any company in your industry. That is brand voice quietly draining away, one reasonable-sounding sentence at a time.
This post shows what that looks like, why it happens, and a practical voice system your team can use so that AI drafts inside your voice instead of replacing it.
Why AI generated content sounds the same
A language model predicts the most likely next words. Ask it to write about “innovative, customer-centric solutions” and it will oblige, because that phrase is statistically everywhere. The more companies ask for the likeliest sentence, the more the same sentence appears.
There is research pointing the same way. In a Science Advances study by Anil Doshi and Oliver Hauser (2024, DOI 10.1126/sciadv.adn5290), 293 writers wrote short stories, some with story ideas from an AI model. Stories written with AI help were rated as more creative and better written, especially for less creative writers. But the AI-assisted stories were also more similar to each other than stories written by humans alone. The authors describe it as a social dilemma: each writer is better off, while the range of content overall narrows.
A caution on scope: that was a controlled experiment on fiction, not on marketing copy. We are not claiming a number for your homepage. But the pattern will be familiar to anyone who has read five competitor sites in a row.
The risk is highest for businesses selling similar things to similar people. If three Pune logistics firms all ask a tool for a “professional, trustworthy About page”, they will get three versions of one page.
Four rewrites: generic versus specific
The companies below are fictional. The “before” lines are the kind of thing AI produces when it is asked for good copy with no further context. The “after” lines are what happens when someone adds a real point of view and real detail.
1. A homepage headline
Fictional company: Kaveri Freight, a Bengaluru logistics firm.
Before: “Delivering excellence in logistics solutions, tailored to your business needs.”
After: “Your Hosur-bound truck leaves by 6 a.m. or we tell you by 5.”
The first line could sit above any freight company in India. The second makes a promise you can check, names a lane, and shows a company confident enough to be held to a time.
2. An About paragraph
Fictional company: Dhanwantri Naturals, an Ayurvedic skincare brand from Kerala.
Before: “At Dhanwantri Naturals, we are passionate about wellness. Our mission is to empower you to live your best life with high-quality, natural products crafted with care.”
After: “My grandmother’s kashayam recipe had four ingredients and one rule: nothing leaves the pot until it is ready. We still work that way. Every batch of our face oil is cold-pressed in Thrissur and tested for 21 days before it goes into a bottle.”
The first paragraph is interchangeable with a thousand wellness brands. The second has a person, a place, a rule and a process. (Fictional, so the figures are illustrative; in your version, every detail must be true and checked.)
3. A product description
Fictional company: LedgerLeaf, accounting software for traders.
Before: “LedgerLeaf is a powerful, all-in-one platform that streamlines your finance operations and unlocks actionable insights.”
After: “LedgerLeaf reads your Tally export on Sunday night and, by Monday morning, tells you which five parties owe you more than 60 days. It drafts the WhatsApp reminder. You press send.”
“Streamlines” and “unlocks” say nothing. The rewrite tells the reader what happens, when, and who is in control at the end.
4. A social post
Fictional company: Chai & Co. Fitness, a gym chain in Hyderabad.
Before: “New year, new you! Start your fitness journey today and unlock your full potential.”
After: “By the third week of January, most memberships go quiet. Ours has a rule: miss two sessions and a trainer calls you, not a bot. Come in Saturday, bring a friend, first session free.”
The first could be posted by anyone, anywhere, in any year. The second reflects how that business actually behaves.
Notice what the “after” versions share. None of them used a smarter model. They used facts only the company had: a lane, a recipe, a Tally export, a trainer’s phone call.
Human in the loop: AI cannot supply what your business knows and its training data does not. Someone senior has to decide what is true, what is distinctive and what the brand will never say. That decision is the voice.
Build a voice system your AI can follow
The fix is not to stop using AI. It is to give it a brief that is specific enough that generic output becomes difficult. We suggest a one-page voice system with four parts.
1. A positioning line
One sentence: who you serve, what you do differently, and why it matters. Not a slogan. A test.
Example (Kaveri Freight): “For manufacturers in the Bengaluru-Hosur belt who cannot afford a missed dispatch, we are the carrier that commits to a departure time and reports honestly when we miss it.”
Every draft should be checkable against it. If a paragraph would be equally true for a competitor, it fails.
2. Three voice traits, each with a do and a don’t
Choose three. More than three and nobody can apply them. Each needs a pair of examples so a new writer, or a model, can tell the difference.
| Trait | Do | Don’t |
|---|---|---|
| Plain | “We ship by 6 a.m.” | “We leverage optimised dispatch workflows.” |
| Specific | “Hosur, Krishnagiri, Salem.” | “Across multiple regions.” |
| Accountable | “If we are late, you hear from us first.” | “We strive for timely delivery.” |
Write your own three from your own customers’ words. Listen to sales calls and WhatsApp chats; the phrases clients use are better than anything a brainstorm will produce.
3. A banned phrases list
This is the most useful part, because it is easy to enforce. Start with phrases that appear on thousands of Indian business pages:
- “In today’s fast-paced world” and “in the ever-evolving landscape of”
- “Tailored solutions” and “end-to-end solutions”
- “Unlock your potential”, “empower”, “elevate”, “seamless”, “cutting-edge”
- “Passionate about”, “customer-centric”, “world-class”
- “Delve into” and “game-changer”
- “Your trusted partner” (unless you can say for how many years, and with whom)
Then add your own, including competitor phrases you do not want to echo. Review the list every quarter; when a new tic starts to show up in drafts, add it.
4. An example bank
Models imitate examples far better than they follow adjectives. Keep 10 to 15 real pieces your company is proud of: a sales email that got a reply, a LinkedIn post a client mentioned, a proposal paragraph, a WhatsApp message from your founder. Label each one with why it works (“names the lane”, “admits the downside”).
Also keep a short facts bank: real numbers, client situations you have permission to mention, prices in rupees (including GST treatment), turnaround times, and things you will not do. This is what stops the model from filling gaps with plausible fiction.
Paste the whole one-page system into every drafting session. Then, ask the model to flag any sentence it wrote that a competitor could use unchanged.
What to do when the draft still reads flat
Even with a good brief, AI drafts will drift towards the middle. Build a short editing pass:
- The swap test. Replace your name with a competitor’s. If the sentence still works, rewrite it.
- The proof test. Every claim needs a number, a name, a place or a visible consequence. If it has none, cut it or find the evidence.
- The ear test. Read it aloud. If you would never say it to a client over chai, change it.
- The privacy and accuracy check. Examples, client details and claims must be real and permitted. Under India’s DPDP Act, personal data in your examples is not yours to paste into just any tool.
Human in the loop: one named person owns the voice. They approve the system, review what goes out under the brand’s name and have the authority to say “this is not us”. Without that owner, AI drafts will pull everything towards average.
How we handle this at Cannyworx
We use AI for first drafts, variations and volume, and a senior person decides what the brand says and signs off every piece before it leaves the studio. Where we build brand and collateral work, the voice system above is written with the client’s team first, because the drafting tools are only as specific as the brief we give them.
Where to start
You do not need a rebrand to fix this. Start with one afternoon:
- Collect your last ten pieces of public copy: homepage, About page, three posts, a proposal intro, two emails.
- Run the swap test on each. Mark every sentence a competitor could use.
- Write your positioning line, pick three traits and draft a first banned list.
- Choose the one person who will own the voice.
The free brand swipe file below gives you before/after rewrites and a fill-in template for each part of the system, so your team can start with a working draft.
Questions people ask
Can AI write in my brand voice?
It can imitate a voice that has been written down clearly, with rules, banned phrases and real examples. It cannot invent one for you, so a person on your team needs to own the voice and edit every draft against it.
How do I stop AI content sounding generic?
Give the AI specifics it cannot guess: your positioning line, real numbers, client situations, product names and a list of phrases it must never use. Then have an editor cut anything that could appear on a competitor's page unchanged.
What is brand voice and why does it matter more with AI?
Brand voice is the consistent way your company sounds: word choice, rhythm, what it will and will not say. When every competitor can generate fluent copy in seconds, a recognisable voice is one of the few things left that makes you look like yourself.
