The short version: Batching three months of content with AI in one sitting looks efficient on paper and dies quietly in the feed. I did it, watched my engagement fall by roughly 40% over six weeks, and went back to writing and posting on the day. The fix isn’t “use less AI”, it’s using it for research and drafting, never for the finished, published thing.
Worth reading next: Your AI Blog Content Is Ranking. It’s Just Not Selling Anything..
The afternoon I “solved” my content problem
Back in February I sat down with ChatGPT, my old topic list, and a spreadsheet, and by 4pm I had 90 days of LinkedIn posts, 12 newsletter drafts, and a stack of Instagram captions ready to schedule. It felt brilliant. I’d been telling clients for years that consistency beats perfection, so I built myself a consistency machine.
I scheduled the lot through a queue tool and told myself I’d freed up two hours a day. For the first fortnight nothing looked wrong. Then I started noticing the comments drying up. A post about “5 AI tools worth paying for” that would normally pull 30 to 40 comments from my usual crowd got two. A newsletter that I’d normally get 15 to 20 replies to got three, and two of those were unsubscribes explaining why.
I went back and reread six weeks of my own scheduled content in one sitting, which is a uncomfortable thing to do. It was fine. Grammatically correct, on brand, hit the word count. It was also nobody. There was no specific client story, no number I’d seen that week, no opinion I’d argue about down the pub. It read like a very competent stranger doing an impression of me.
The bit nobody selling you an AI content calendar wants to say out loud
Here’s the uncomfortable truth: batching content with AI doesn’t just risk sounding generic, it structurally guarantees it, because AI has no idea what happened to you yesterday. It can’t reference the client call where a business owner told you she’d rather close her shop than learn one more piece of software. It doesn’t know the LinkedIn post that blew up last Tuesday or the industry news that dropped this morning. Everything it writes for a date three months out is written blind to that date. You are trading relevance, which is the actual thing that makes social content work, for volume, which nobody asked you for.
The people selling “generate 90 days of content in an hour” as a feature are not lying about the mechanics. It does generate 90 days of content in an hour. They’re just quiet about what that content is worth once it’s live, which in my case was measurably less than what I was writing badly and quickly on the day.
What happened to the numbers
- LinkedIn post engagement (comments plus shares) dropped from a six month average of around 35 per post to roughly 20 during the scheduled batch, a fall of close to 40%
- Newsletter reply rate went from around 4% of openers replying to under 1%
- Two long-standing subscribers unsubscribed with the specific feedback that the emails “felt like a template”
- My own time saved was real but smaller than promised: roughly 6 hours a week, not the “10 hours back” the tool’s marketing page claimed
That last point matters more than people admit. The tools do save time. They just don’t save as much as the sales page says once you factor in the editing, the fact-checking, and the inevitable rewriting when a scheduled post turns out to reference something that’s no longer true.
What I do instead, step by step
I didn’t ditch AI. I ditched the batching and the scheduling of finished posts. This is the actual workflow I run now, most days it takes about 30 to 40 minutes total across all channels:
- Step 1, five minutes: I ask AI (usually Claude or ChatGPT) to summarise what’s happening in my niche that week, based on articles or LinkedIn posts I paste in myself. This is research, not writing.
- Step 2, ten minutes: I write the actual post or email myself, first draft, badly, including one specific thing that happened to me or a client that week. A number, a name (with permission), a mistake I made.
- Step 3, five minutes: I ask AI to tighten the structure and check for waffle, not to add ideas. I reject most of its suggestions.
- Step 4, five minutes: I post it the same day it’s written, or the next morning at the latest. Nothing sits in a queue for more than 48 hours.
- Step 5: On newsletter days I follow the same process but I lean on what I’ve learned running mine through Mailchimp’s approach to segmented, personal email, which is built entirely around emails that feel like they came from a person that week, not a content bank.
The only thing I still batch is topic ideas, a running list of 20 to 30 headlines I might write about. Ideas keep, first drafts don’t.
Why this matters more on Instagram and LinkedIn than anywhere else
Text and image platforms are the worst place to run pre-written AI content because the algorithms on both are actively rewarding recency and reaction speed. LinkedIn’s own engineers have said publicly that dwell time and early comment velocity are ranking signals. A post written three months ago about a trend that’s since moved on doesn’t just underperform, it can actively work against your account’s reach on the next post too.
Compare that with how the strongest personal brands operate. Look at how Instagram-native brands build genuine engagement, it’s rarely from perfectly scheduled content, it’s from responding inside the news cycle of their own industry within hours, not months. Airbnb’s marketing team does the same thing on a bigger budget, they react to what’s happening in travel and culture that week rather than running a fixed calendar, which is one of the reasons Airbnb’s marketing strategy still feels current a decade after they built the playbook.
Where AI earns its place instead
I’m not anti-AI, I use it every single day, just not for the final published word. Here’s where it’s earned its keep in my business this year:
- Research and pattern spotting. Feeding it a month of my own analytics and asking what topics did well, which it’s far faster at than me trawling spreadsheets.
- First-pass structure. Turning a voice note I’ve recorded in the car into a rough outline I then rewrite.
- Interactive tools, not static posts. Instead of another AI-written blog post, I built an interactive calculator that visitors fill in themselves, which converts far better than a passive article because the person is doing something, not just reading something AI generated for them.
- Product and offer thinking. AI is useful for stress-testing an offer before you launch it, in the same spirit as the pricing and value-stacking principles in Alex Hormozi’s business lessons, ask it to poke holes in your offer, then decide yourself which holes are real.
A quick sanity check before you schedule anything AI wrote
Ask yourself these before you queue a post more than two days out:
- Does it reference something that only makes sense this week?
- Would I say this out loud to a client on a call, in these exact words?
- Is there a real number, name, or mistake in it that AI could not have invented?
- If I read this alongside ten other posts from my industry, could I tell it was mine without the logo?
If the answer to any of those is no, it goes back in the drafts folder, not the queue. That one rule has done more for my engagement than any tool I bought this year.
The honest cost of doing it my way
I won’t pretend this is free. Writing live, every day, takes discipline that batching was specifically designed to remove. There are weeks I miss a day because a client call ran long or my son needed picking up early. Those gaps used to feel like failure when I was chasing a perfect 90 day calendar. Now they don’t, because an empty Tuesday is less damaging to a personal brand than a Tuesday filled with something that sounds like nobody in particular.
If you run a small business and you’re weighing up whether to batch a quarter of content with AI this month, my honest advice after doing exactly that is: don’t. Batch your ideas, batch your research, batch your topic list. Write and post the actual words close to the day they go live, every time.
Frequently asked questions
Is it bad to schedule AI-generated content in advance?
Scheduling itself is fine, it’s the gap between writing and publishing that causes the problem. If AI writes a post today and it goes live today, that’s low risk. If AI writes it in January for a March publish date, it’s written blind to whatever’s happening in March, and readers can feel that even if they can’t name why.
How much time does writing content live cost compared to batching?
In my own case, batching saved roughly 10 hours a month more than writing live, but cost me a 40% drop in engagement and two lost subscribers with specific feedback about it feeling templated. The time saved wasn’t worth the reach lost, especially since reach is what turns content into enquiries.
Can AI ever be trusted to write a finished social media post?
It can write a finished draft you then rewrite in your own words, with your own specific example added in. Trusting it to write the final, published version without a real detail from your week added by you is where the generic tone creeps in, every single time I’ve tested it.
What should small business owners batch instead of finished content?
Batch ideas, topic lists, and research summaries, never the final wording. A list of 20 to 30 headlines takes an afternoon and stays useful for months, because an idea doesn’t go stale the way a fully written, dated post does.
Related reading: The AI Notetaker Sitting In On Your Client Calls Might Be Breaking Your NDA and AI Meeting Notetakers Are Quietly Costing You Client Trust.
