Check six things before it goes live: every fact, number, name and date against a source; that it sounds like you and avoids your banned phrases; platform mechanics (length, hashtags, tags, links, alt text); rights and labels (music, footage, people, AI-media disclosure); context on the day it posts; and who approved it. Most posts take three to five minutes.
The group teams skip most is context. AI drafts posts days ahead and scheduling tools publish them on time regardless of what's happening, so a cheerful post can land on the day a client announces redundancies or your industry is dealing with bad news. A person looking at the queue that morning is the only check that catches it.
Facts, figures and names
AI fills social posts with specifics because specifics read well. That's also why they're the first thing to check. For a small video production company posting its own work and clients' films, this group catches most of the serious errors:
- Every number has a source. Why: "50,000 views in the first week" is exactly the kind of figure a model invents to make a post sound successful. How: find the number in the analytics, the client's report or your records; if you can't, cut it.
- Names are spelt as their owners spell them. Why: client names, product names and people's names are easy for AI to "correct" into something plausible. How: copy names from the client's own website or signed paperwork, never from memory.
- Job titles are current. Why: people change roles, and AI often uses whatever title appeared most in older material. How: check the person's current profile or ask them.
- Dates and times are right, including the day of the week. Why: "this Thursday, 9 October" is wrong if the 9th is a Friday, and AI gets weekdays wrong more often than you'd expect. How: check against a calendar, and write out event times with the time zone if your audience is spread out.
- Claims have evidence. Why: "award-winning", "the best in the business" (or any superlative) and "trusted by" are claims, and a public post is advertising. How: check the claim against your evidence file; backing up every claim AI writes explains what counts.
Voice, tone and the phrases that give AI away
A post can be accurate and still read like it came from a machine. Followers notice faster than you might think, and clients whose accounts you manage notice faster still.
- It sounds like the account. Why: each account has a voice, and AI drifts towards a generic upbeat one. How: read it aloud next to the last three posts on the account; if it sounds like a different person, rewrite the first line.
- No banned phrases. Why: every brand has phrases it never uses, and AI favourites such as "exciting news", "we're thrilled to announce" and "a true labour of love" make a feed look automated. How: search the draft against your list; a banned-phrase list for your brand shows how to build one.
- No invented enthusiasm about other people. Why: "our amazing client was blown away" puts words and feelings in someone else's mouth. How: describe what happened, and quote the client only with their approval.
- The opening line works on its own. Why: most platforms show only the first line or two before a "more" cut-off. How: read just the preview in the scheduler; if it says nothing specific, move the most interesting fact up.
- Emojis and exclamation marks match the account's habits. Why: a sudden burst of both is one of the quickest tells of an AI draft. How: compare with recent posts and trim to the account's norm.
Platform mechanics that break silently
These are the errors that don't look like errors in a draft document and only show once the post is live. Each platform has its own limits, and they change, so the person who owns social checks them every quarter. As checked in September 2026:
- Length. LinkedIn posts are capped at 3,000 characters. Standard posts on X are limited to 280 characters; longer posts, up to 25,000 characters, need a Premium subscription, and anyone you mention after the first 280 characters currently isn't notified. How: paste the final text into the scheduler's preview for each platform, not just the document.
- Hashtags. Instagram allows no more than five hashtags on a post, a cap in place since December 2025, yet AI drafts still routinely suggest 15 or 20. How: count them for every Instagram post, and decide which five earn their place.
- Tags and handles. Why: AI guesses handles from company names, and the guess can belong to a different business or an abandoned account. How: open every @handle in the app before posting and confirm it's the right account.
- Links. Why: a link to the wrong page, a staging site or an expired event page wastes the post. How: click every link from the scheduled version, and check tracking parameters are on the link rather than in the visible text.
- Alt text and captions. Why: screen-reader users rely on alt text, and many people watch video with the sound off. How: write alt text that describes the image rather than the prompt that made it, and check that burned-in or uploaded captions match the final edit.
- Search visibility. Why: public posts from professional Instagram accounts can appear in Google and Bing results since July 2025, so a wrong figure can outlive the feed. How: treat captions with the same care as a web page.
Rights, consent and AI labels
For a video company this group matters more than for most, because every post contains footage, music and usually people.
- The client has approved this post, not just the film. Why: approving a finished film isn't the same as agreeing to your announcing it, today, with those stills. How: send the exact post, with media, for written approval.
- Music is licensed for this use. Why: a track licensed for the client's own channels may not cover your showreel post. How: check the licence terms for the platform and the account.
- People in shot have agreed. Why: release forms often cover the client's use but not the production company's promotion. How: check the release wording before using faces in your own posts.
- AI-generated or altered media is labelled where required. Why: platforms require it. Meta asks for a disclosure whenever you post photorealistic video, or audio that sounds real, that was made or changed digitally. TikTok requires a label on AI-generated content showing realistic scenes or people. YouTube asks creators to use its "AI use" setting for realistic altered or synthetic content, though not for AI help with scripts, titles, thumbnails or captions. How: note in the draft whether any image, voice or footage was generated or altered, and set the label when posting. The detail is in the platform-by-platform labelling guide.
- Generated images have been checked for errors. Why: extra fingers, garbled signage and impossible reflections are still common. How: look at them at full size; checking AI-generated images for errors has a full list.
Timing and context on the day it posts
This is the only group that can't be finished when the post is drafted. It happens on the morning of publication, for everything scheduled that day.
- Nothing in the news makes it tone-deaf. Why: a jokey post about "explosive" results reads very differently on a day of bad news. How: a two-minute scan of the headlines and your sector's news before the queue runs.
- The client's situation hasn't changed. Why: a celebratory post about a client on the day they announce job cuts or a product recall damages both of you. How: check the client's own channels and news that morning; for client posts, confirm nothing has changed since approval.
- Embargoes and launch dates still hold. Why: launches move, and a scheduled post doesn't know. How: confirm the date with the client for any post tied to their launch.
- It isn't a duplicate. Why: the same announcement drafted twice by AI, scheduled by two people, looks careless. How: scan the week's queue for the same topic.
Who signs off, and how fast
A social media approval process should match the risk of each post, or it becomes either a bottleneck or a formality. The video company's rules:
| Post type | Checks | Approver | Target turnaround |
|---|---|---|---|
| Own account, no claims or people (behind-the-scenes, tips) | Checklist by the drafter | None; weekly spot check | Same day |
| Own account featuring client work | Full checklist | Producer, plus the client's written OK | 2 working days |
| Posts on a client's account | Full checklist | The client, on the exact post and media | Agreed in the contract |
| Any post with a figure, claim or offer | Full checklist plus evidence | Producer | 1 working day |
| Replies during a complaint or crisis | No AI drafting | Owner writes | Same hour |
Many scheduling tools offer an approval step on team plans; check whether yours does before building a workaround in email. The same lanes-and-sign-off logic applies to longer content, set out in full in the AI content approval workflow.
A drafting prompt that makes checking faster
The checklist goes quicker when the draft arrives with its working shown. Instead of asking for a finished post, ask for the post plus the material the checker needs:
Draft an Instagram post and a LinkedIn post announcing the film
described below. Use only the facts in the brief.
Then list, under separate headings:
FACTS USED: each number, name and date, with where it came from
HANDLES: every account to tag, marked [CONFIRM] - don't guess
HASHTAGS: at most 5 for Instagram
ALT TEXT: one sentence describing the cover image
AI MEDIA: say whether any image, audio or footage in the brief was
generated or altered by AI
If the brief doesn't give a figure you'd normally include, leave it
out rather than estimating.
[brief: client, film, crew size, shoot days, release date, media]An illustrative extract of what comes back:
FACTS USED: crew of seven (brief); three shoot days (brief);
release at the client's launch event, Thursday (brief)
HANDLES: [client] [CONFIRM]; our cinematographer [CONFIRM]
HASHTAGS: #videoproduction #brandfilm #corporatevideo
ALT TEXT: A furniture maker planing an oak tabletop in a sunlit
workshop, sawdust in the air.
AI MEDIA: None stated in the brief.That extract turns half the checklist into ticking boxes. What still needs a person: the [CONFIRM] handles must be opened in the app, "None stated" on AI media is only as good as the brief, and the Thursday date needs confirming with the client on the morning it posts. The draft is also more honest by design, because the prompt told the AI to leave a figure out rather than invent one.
One AI-drafted post, checked line by line
Here's an illustrative Instagram draft the company's AI assistant produced to announce a finished film for a client, a furniture maker:
So proud to unveil our latest film for [client]! This award-winning campaign has already racked up 50,000 views in its first week. Huge thanks to the amazing team at @[guessed handle]. Shot on location over three days with a crew of 12. #videoproduction #filmmaking #corporatevideo #brandfilm #behindthescenes #cinematography #storytelling #marketing #videomarketing #contentcreation #furniture #4k
The checklist found seven problems in about four minutes:
- "Award-winning": the film hasn't won anything. Cut.
- "50,000 views in its first week": the film isn't public yet, so there are no views. Invented.
- @[guessed handle]: belongs to a different company with a similar name. The client's real handle has an underscore.
- "A crew of 12": it was seven.
- Twelve hashtags: over Instagram's cap of five.
- Timing: the client's launch event is on Thursday, and the film is embargoed until then. Scheduled for Tuesday, the post would have scooped the client's own announcement.
- Opening line: "So proud to unveil" is the company's most-used AI opener and says nothing specific.
The version that went out on Thursday afternoon, after the client's launch and with the client's written OK:
Three days, a crew of seven, and a very patient furniture maker. Our new film for [client] went live at their launch this morning; watch it on their channel (link in bio). #videoproduction #brandfilm #corporatevideo
With alt text on the cover image describing the workshop scene, the right handle tagged in the image, and no AI label needed because the footage was all filmed, not generated.
When a post goes out wrong anyway
Something will get through eventually. What matters then is speed and openness:
- Decide in minutes whether to edit or delete. A wrong figure or a typo in a post with little engagement can simply be corrected. A post that is offensive, breaches an embargo or tags the wrong company should come down, because every hour it stays up adds screenshots.
- Don't delete quietly if people have engaged. If there are comments or shares, a short correction ("An earlier version of this post gave the wrong crew size; it was seven") keeps trust better than a post that vanishes.
- Tell the client first if the post was about them or on their account, before they hear it from someone else.
- Check where else it went. Cross-posted versions, scheduled repeats and newsletter embeds often carry the same error.
- Log it and fix the checklist. Every miss should end with a new line in the checklist or the drafting prompt, so it can't happen the same way twice.
A week of posts: time and catches
In an illustrative week, the company publishes about 15 posts: six on its own accounts and nine across three clients' accounts. At three to five minutes per post for the checklist, plus ten minutes each morning for the timing check, the total is roughly 1 hour 50 minutes a week. At an internal rate of $50 an hour, about $90.
That week's catches: two invented figures, one wrong handle, four Instagram posts over the hashtag cap, one client post whose launch had moved, and one AI-generated background in a graphic that needed a label. Any one of the client-side misses would have cost more than a month of checking in goodwill alone. And the time falls as the lists improve: once the banned-phrase list and the handle list are complete, the voice and tagging checks take seconds.
Further reads
- How to Create On-Brand Social Media Graphics With AI — Create social graphics with AI that match your brand.
- How to Write Instagram Captions With AI That Sound Like You — Draft Instagram captions with AI that sound like you.
- How to Write LinkedIn Posts With AI Without Sounding Like AI — Write LinkedIn posts with AI that don't read as AI.
- Can AI Run Your Social Media on Autopilot? What Breaks — What breaks when social posting runs without checks.
- How to Keep AI-Made Videos On-Brand: Captions, Voice and Intros — Keep AI-made video clips on-brand before they reach a post.
- Email Newsletter QA Checklist Before You Hit Send — The matching checklist for email sends.
- AI Prompts for Bakers: Pricing, Captions and Customer Replies — Twelve copy-ready prompts for bakers covering costing and prices, social captions, and customer replies, each with what to check before you use the answer.
- How Photographers Can Fill Mini-Session Slots With AI Marketing — A three-week, waitlist-first campaign for mini sessions: slot maths, past-client segments, copy prompts and honest ways to fill the last slots.
- How to Plan a Month of Barbershop Instagram Content With AI — A one-hour planning session, a sample four-week grid, a shot list for filming the lot in one afternoon, and captions that sound like your shop.
- How to Create Outfit and Styling Posts for a Boutique With AI — Turn a boutique's stock list into weekly outfit carousels: AI plans combinations and writes the copy, while the photos stay real and the sizes stay honest.
- How Food Trucks Can Use AI to Plan Stock, Pitches and Posts — A trading log, a weekly planning routine and copyable prompts that help a food truck plan prep, choose pitches and post its schedule with AI.
- How Florists Can Use AI for Orders, Reviews and Social Posts — Worked examples for a flower shop: a rushed phone order turned into a docket, a substitution message, a year of reviews analysed, and accurate captions.
- Turning One Online Course Into Emails and Posts With AI — Transcribe the lessons, pull out one idea, mistake and quick win per lesson, and turn each into an email or post. Twenty lessons can fill two months.
- Charity Appeals and Social Posts With AI That Keep Your Voice — How a small charity turns past appeals into a voice card, briefs AI with a real consented story, and edits drafts until they sound like itself.
- Marketing a Physio Clinic With AI: Posts, Tips, and Newsletters — A physio clinic's marketing routine built on AI: caseload-led topics, voice notes turned into posts and newsletters, claims checks and booking tracking.
- How Much Does It Cost to Automate Social Media With AI? — Scheduler, AI writing, design, automation and DM costs itemised, three budgets built line by line, and the review hours no tool removes.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: LinkedIn Help (post character limit); X help pages (post length, longer posts, mention notifications); Meta newsroom and transparency pages (AI labels for photorealistic video and realistic audio); TikTok support pages (AI-generated content labels); YouTube Help (AI use disclosure setting); our verified fact sheet (Instagram hashtag cap, Instagram posts in search results).