A marketing manager at an outdoor gear company told us they replaced their freelance copywriter with an AI tool. The results were immediate. Blog posts that used to take three days now took thirty minutes. Email campaigns went from one a week to four. Social posts were written in bulk a month ahead of time.
Six months later, their email open rates had dropped 22%. Social engagement was flat. Organic traffic had actually increased, but time on page was down across the board. People were landing on the site and leaving faster than before.
The copy was not bad. It was grammatically correct, well-structured, and hit the right keywords. The problem was that it sounded exactly like every other AI-generated marketing copy on the internet. And their audience noticed.
The Sameness Problem
Run the same prompt through any major AI model and you get copy that follows predictable patterns. Short punchy opener. Three supporting points. Call to action. The structure works. It works so well that millions of businesses are now using it.
That is the problem. When everyone's marketing copy follows the same patterns, uses the same sentence rhythms, and reaches for the same emotional beats, nothing stands out. The copy is fine. Fine does not get remembered.
This matters more for some businesses than others. A plumbing company probably does not need a distinctive brand voice in their Google Ads. But a DTC outdoor brand competing against 400 other companies selling merino wool base layers? Voice is the differentiator.
The outdoor gear company had spent years building a voice that was irreverent, specific, and packed with references their audience recognized. Their copywriter had climbed half the peaks in the Cascades. That showed up in the writing. When the AI took over, the specificity disappeared. The voice became generic.
What AI Copy Actually Does Well
This is not an argument against using AI for any marketing copy. AI is good at specific tasks.
First drafts. Starting from a blank page is the hardest part of writing. AI eliminates that friction. A rough draft you can edit is faster than staring at a cursor.
Variations. Need to test five subject lines for an email campaign? AI can generate twenty in a minute. You pick the best ones.
Reformatting. Turn a blog post into a LinkedIn post, an email teaser, and a Twitter thread. AI handles the conversion well because the original thinking is already done.
Boilerplate. Terms of service, privacy policies, FAQ answers that are factual rather than voice-driven. AI produces these quickly and accurately enough.
The pattern: AI is most useful when the thinking has already been done by a human. The strategy, the positioning, the specific claims. AI can take that thinking and put it into formats quickly. The trouble starts when you hand AI both the thinking and the writing.
The Voice Gap
Brand voice is built on specifics that AI does not have access to.
Your copywriter knows that your customers hate the word "adventure" because every competitor uses it. They know that your audience responds to gear specs but not aspirational lifestyle shots. They know that your CEO has a thing about Oxford commas and will reject any copy that includes them.
AI does not know any of this. You can put some of it into a prompt. "Write in a casual, outdoorsy tone. Avoid cliches. Our audience is experienced backcountry skiers, not beginners." The output will be better than a generic prompt. But it will not be the same as copy from someone who understands the brand from the inside.
The gap is subtle but cumulative. One AI-generated blog post is fine. A hundred of them, published over six months, erodes the voice that made the brand recognizable.
The SEO Trap
A common argument for AI copy is that it helps with SEO volume. More pages indexed, more keywords covered, more chances to rank.
This worked for a while. It is working less well now.
Google has been clear about its position on AI-generated content. The content itself is not penalized for being AI-generated. But thin, repetitive, low-value content is penalized regardless of who or what wrote it. And AI-generated content at scale tends to be thin, repetitive, and low-value because nobody is editing it.
The businesses that are winning with AI content are the ones that use AI to produce first drafts and then invest human time in making those drafts specific, accurate, and useful. The businesses that are losing are the ones that publish the first draft as-is and move on to the next one.
More content is only better than less content if the content is good. Fifty mediocre blog posts do not outrank ten excellent ones.
The Real Cost Calculation
The freelance copywriter cost $4,000 a month. The AI tool costs $100. On paper, that is a $47,000 annual saving.
But the outdoor gear company's email revenue dropped when open rates fell. Social engagement declined. Brand recall in their customer surveys went down for the first time in four years.
They could not put a precise number on the revenue impact. Marketing attribution is messy. But the marketing director told us the "savings" from cutting the copywriter probably cost them more in the second half of the year than the copywriter's full annual salary.
They rehired a copywriter. Not the same one. She now works alongside the AI tools. She does the strategic thinking, the brand voice work, and the editing. AI does the first drafts and variations. The combination produces more content than either one alone, and the content sounds like the brand again.
Where This Leaves You
If you are a marketing or product owner evaluating AI for content, the question is not "can AI write this?" It can write almost anything. The question is "will AI-written copy be good enough to represent this brand to people who have choices?"
For some content, the answer is yes. Product descriptions that are primarily factual. Internal documentation. Email transactional copy. Ad variations for testing.
For content that carries your brand voice, the answer is almost always no. Not without human involvement. The AI draft is a starting point. The human editor turns it into something that sounds like your company instead of every company.
The savings from AI copy are real. But they are smaller than they look on a spreadsheet, because the spreadsheet does not have a line item for "sounds like everyone else."