The short version: Batch-writing a month of AI content in one sitting feels efficient, but it flattens your voice into one long average tone that both readers and the platform’s algorithm can spot within a few posts. I tested it on my own LinkedIn for three weeks, my reach dropped by more than three quarters, and I went back to writing daily with AI only doing the first draft. The fix isn’t “use AI less”, it’s using it differently, one post at a time, with something real dropped into every single one.
The afternoon I wrote a month of content in three hours
Last spring, mid-rebuild of my business, I decided to be efficient. I sat down with ChatGPT, fed it ten topics, and by the end of a Tuesday afternoon I had 30 LinkedIn posts drafted, polished and loaded into my scheduler. Three hours of work covering four weeks of content. I felt smug about it. I’d finally cracked the thing every productivity guru promises: content on autopilot.
I want to be clear about what I did, because it’s exactly what most small business owners try when they discover AI writing tools. I gave it my usual topics: AI adoption for small teams, working from home, rebuilding after a rough patch. I told it my tone. I asked for hooks, short lines, a call to action. It did all of that competently. Grammatically the posts were fine. Structurally they followed every rule I usually teach clients about short paragraphs and strong openers.
They just weren’t mine, and it turned out that mattered a lot more than I expected.
What happened to my numbers over the next three weeks
My average LinkedIn post normally pulls somewhere between 3,500 and 5,000 impressions and an engagement rate around 3.5 to 4 percent, comments included. That’s not viral by anyone’s standard, but it’s steady, and it’s mine.
Within the first week of the batched posts going out, impressions dropped to around 1,800. By week two they were down to 900 on some days. Engagement rate fell to 0.6 percent. Comments, which are usually where I do most of my actual selling in the DMs afterwards, almost disappeared. I went from ten or twelve meaningful conversations a week off the back of posts to two.
The posts weren’t bad. That’s the uncomfortable part. If you read any single one of them in isolation, it looked fine. But read five in a row and you could feel the sameness, the same rhythm of short punchy line, slightly longer line, rhetorical question, list, call to action. Human readers pick that up faster than we give them credit for. Several people who’d been reading me for years told me, gently, that something felt “off” that month. Nobody could say what. They just scrolled past.
The bit the AI content courses don’t want you thinking about
There’s a whole industry now selling you the promise of a month of content in an afternoon, and a lot of it is built by people who make money the moment you buy the course, not the moment your engagement holds up three weeks later. Nobody selling “30 days of content in 30 minutes” is showing you the impressions graph a month out. I’ve looked. They show you the time saved on day one and never circle back.
Here’s the part that’s hard to sit with: the efficiency is real, but it’s efficiency for you, not for your reader. Batching content optimises your calendar. It does nothing for the person on the other end scrolling their feed, and they are the only ones whose behaviour decides whether your business gets found. Companies that get this right, like Hotjar’s marketing team, build their content around what they can see users doing, not around what’s easiest to produce on a Tuesday afternoon.
I’m not against AI writing. I use it every day. I’m against pretending that removing yourself from the process for four weeks at a time has no cost, because it does, and the cost shows up in the metrics, not in the finished post.
Why the algorithm can smell a template
LinkedIn, Instagram and every other platform are built to reward content that keeps people on the platform longer, and they measure this through dwell time, replies, and shares, not likes. Content generated in one long session tends to share vocabulary, sentence length and structure across all 30 pieces, because it comes from the same short prompt conversation with the same model settings. The platform doesn’t need to “detect AI” to penalise this. It just needs to notice that engagement per post is dropping, and it quietly shows the next post to fewer people. That’s the whole mechanism. No conspiracy needed, just a feedback loop doing exactly what it’s designed to do.
This is the same reason a brand like Revolut varies its content format constantly rather than running the same template on repeat, and why Pinterest’s own content strategy leans so heavily on fresh, specific visual detail rather than repeating a formula. Sameness is the one thing every platform’s ranking system is quietly built to punish, even when a human reader can’t immediately explain why they’ve stopped clicking.
What I do instead now
I still use AI for a first draft. I have not gone back to writing everything from a blank page, because that’s not realistic for anyone running a business on their own. But the process changed completely.
- I keep a running note on my phone of one real thing that happened that day: a client comment, a DM that made me laugh, a number from a campaign, a mistake I made. That’s the seed of the post, never the AI prompt.
- I write the opening line myself, every time, before I touch a tool. It’s usually the worst-written line in the post and also the most human one, which is exactly the point.
- I let AI build the middle structure if I want it to, the list, the flow, the transitions, but I always change at least one specific detail per post that could only be true of my week, not any week.
- I post close to real time rather than scheduling a month out, so I can still reply to comments within an hour or two while the post is live.
- Once a week I check the analytics rather than glancing at likes, specifically dwell time and comment quality, not comment count.
This costs me more time than batching, roughly 40 minutes a day instead of three hours a month. But my engagement rate is back above 3 percent and DMs off the back of posts are back to double figures a week. The time cost is real. So is the return.
A quick way to test if your content has gone stale
Here’s a test I now run on my own content every few weeks, and I’d recommend it to any small business owner posting regularly. Pull your last five posts and read them back to back, out loud, in one sitting. If you can’t tell which day you wrote each one without checking the date, your readers can’t either, and neither can the algorithm distinguish them as anything worth showing to new people. It’s the content version of the interactive tools I talk about with clients when I explain why a calculator or quiz outperforms a static blog post: the whole value is in something that responds specifically to the person using it, not a fixed template dressed up five different ways.
Another quick check: ask someone who knows your business, a business partner, your accountant, a client, to read three of your recent posts and tell you one thing they learned about your week specifically. If they can’t answer, the post could belong to anyone in your industry, and that’s the actual problem batching creates. It’s not that AI wrote it. It’s that nothing in it belonged only to you.
When batching still makes sense
I’m not telling anyone to abandon efficiency entirely. There are places AI batching is still the right call:
- Evergreen how-to content that doesn’t date, like a help centre article or a product FAQ page.
- Repurposing one long piece, like a webinar transcript, into several shorter posts, since these come from a single real source anyway.
- Internal documentation, onboarding material, and process guides your team uses, where consistency is the goal, not connection.
What doesn’t work is applying the same batching logic to any channel where the whole point is relationship, comments, and trust built over time. That’s LinkedIn, that’s your email newsletter, that’s Instagram captions. Those channels are supposed to feel like they came from a person that week, because that’s the entire reason someone follows a business account instead of just reading its website.
If you’re building this out for the first time and want someone to sit with you and design a content system that still holds up three months from now rather than three days, this is exactly the kind of gap an AI consultant for a small business should be closing with you, not writing a template and disappearing. It’s worth remembering, too, that the businesses whose strategies get studied and copied, like Miro, built their reputation on specificity over years, not on a month of content produced in one sitting. And some of the best business advice on this exact point is a century old rather than new: Dale Carnegie was making the case for genuine attention over generic charm long before anyone had heard of a large language model, a point worth revisiting in his business lessons if you want the older version of the same argument.
Related reading: ai content sameness small business.
Frequently asked questions
Does AI-written content get penalised by LinkedIn or Instagram?
Neither platform confirms it detects AI directly, but both rank content on engagement signals like dwell time, comments, and shares. Batched AI content tends to share structure and vocabulary across posts, which lowers those signals over time, and the platform responds by showing it to fewer people. The effect looks identical to a penalty even without one being explicitly applied.
How much time should I spend on a single AI-assisted post?
Around 30 to 45 minutes works well for most small business owners: five minutes noting the real detail from your day, ten minutes letting AI build structure, and the rest rewriting the opening line and one section yourself so it reads as specifically yours.
Is it ever fine to schedule a full month of posts at once?
Yes, for evergreen or repurposed content such as help articles or clips from a webinar. It’s the relationship-building channels, LinkedIn, newsletters, Instagram captions, where scheduling a month ahead tends to flatten the tone and hurt reach.
What’s the fastest way to tell if my content has started sounding generic?
Read your last five posts back to back in one sitting. If you can’t tell which day each one came from without checking the date, neither can your reader, and that sameness is exactly what slows your reach down.
Related reading: Why Instagram Reach Dropped and How to Fix It in 2026 and How to Find Out Who Viewed Your Instagram Profile (What Happens When You Try).
For the bigger picture, see my full guide to AI marketing.
