A month of social posts in two hours — without sounding like a robot.

Two hours, once a month, produces 30 days of platform-specific social posts that don't sound like AI wrote them. The trick is the order: collect raw material first, run the right prompt second, and apply the editorial filter third. Most owners skip step three and that's why their AI posts read like AI posts.

The biggest reason small business owners can't keep social media going is the volume problem. You need 8–12 posts a month per channel. Two channels minimum. That's 16–24 posts you have to write while also running your business. Most owners give up by week three of any social media plan they make on January 1st.

AI fixes the volume problem. It doesn't fix the sounding like AI problem — and that's where most owners crash. They paste their AI output directly to LinkedIn and learn within a week that audiences smell AI from a mile away. The drop-off in engagement is severe enough that "no posts" is sometimes better than "obviously-AI posts."

This guide is the workflow we use to get both: the volume of AI and the voice of a human. It takes about two hours, total, once a month.

AI fixes the volume problem. It doesn't fix the “sounding like AI” problem. The editorial filter is what bridges the gap.

The recipe

Step 1 — Pull your raw material

This is the step most owners skip, and it's why their AI posts sound generic. The AI can't write specific posts about your business if you give it nothing specific to work with. Garbage in, generic out.

Open a blank doc. For 20 minutes, write down:

• Five specific things that happened in your business this month. Not "we had a great month." A specific moment. A specific customer. A specific story. The longer the receipt is, the better the posts will be.
• Three questions customers asked you this month. The actual words they used. These become FAQ-style posts.
• Two things you fixed or improved. A new service, a process change, a tool you switched. These become "what's new" posts.
• One unexpected story or moment. The thing you'd tell a friend at dinner. This becomes the post that actually gets engagement.

This 20-minute step is the difference between a month of social posts that sound like you and a month that sounds like every other AI-driven brand on Instagram.

Step 2 — Run the master prompt

Now paste the template (below) into your AI tool. Replace the bracketed sections with your raw material from Step 1, your platform, and a couple of voice notes.

The output should be 12–15 post drafts. Don't try to get all 30 in this step — overproducing means more editing time and worse signal. Aim for 12–15 good ones, edit them in step 3, then run the prompt a second time if you need more.

Step 3 — Apply the editorial filter

This is the step that separates posts-that-sound-like-you from posts-that-sound-like-AI. You're going to read each draft and apply a four-part filter.

Filter 1 — Read it out loud. Anything that sounds like nobody you know would actually say it gets rewritten or cut. The mouth test is more reliable than any other AI-tell heuristic.

Filter 2 — Kill the banned phrases. Specifically: "leverage," "harness," "in today's world," "game-changer," "unlock the power of," "revolutionize," "elevate," "delve into," "embark on," "tapestry," and any sentence that starts with "Imagine if…" These are the AI fingerprints that audiences recognize instantly.

Filter 3 — Add specifics. If the AI wrote "we had a great month for our team" — rewrite to "Sarah closed her first three solo cases this month." Specifics are what make posts feel like a person wrote them.

Filter 4 — Cut the openers. AI loves openers like "Did you know…" and "Here's the truth about…" Cut them. Start the post on the second sentence.

After four filters, expect to keep ~80% of the drafts. The other 20% were either repetitive or unfixable. That's normal.

Step 4 — Schedule the month

Drop the polished posts into Buffer free, your CRM's built-in scheduler, or whatever you already use. Don't post every day — 3x a week per channel is enough. Consistency beats volume.

Mix post types so the feed doesn't read like a content factory. We use a rough 4-1-1 ratio: four "value" posts (tips, stories, behind-the-scenes), one "ask" post (question to audience), one "promotional" post (offer or service mention). Repeat weekly.

The master prompt template

Save this. Paste it into Claude, ChatGPT, or Gemini. Replace the bracketed sections with your real raw material from Step 1.

That's the whole prompt. Adjust the tone words and the banned-phrase list to match your specific voice. Save the prompt as a doc and run it once a month with new raw material.

Spotting AI-tells

These are the patterns that tell an audience "AI wrote this":

The first one could be about any business in any industry. The second one is unmistakably about this business and this week. Specificity is what AI tools can't generate without your input — and it's what audiences respond to.

Common mistakes

1. Skipping Step 1 because it feels like work.

The 20 minutes of raw material is the entire reason this works. Skip it and you'll spend hours editing generic posts into something usable — net more time, worse output. Don't skip.

2. Posting drafts directly without editing.

Even the best prompt produces drafts, not finals. The 45-minute editing step is non-negotiable. Posts that go up unedited tank engagement and train your audience to scroll past your content.

3. Trying to do all 30 posts at once.

15 polished posts is plenty for a month at 3x/week per channel on one platform. If you want a second channel, run the prompt again with the platform field swapped — most posts will need rewriting for the platform's voice anyway.

4. Using ChatGPT's most generic settings.

The free tiers of ChatGPT, Claude, and Gemini all work for this. But the quality difference between vanilla settings and a saved system prompt with your tone notes is significant. Set up a saved instruction set or custom GPT for this workflow once, then reuse it forever.

That's the whole workflow. Two hours a month. 30 days of social posts that sound like a person wrote them. The first time through is slow because you're learning the prompt; by month three you'll be running it on autopilot.

If you want to know whether social is even your highest-leverage channel right now — versus reactivation, reviews, or local SEO — the free scan ranks them. Sixty seconds, no credit card.