Social media changes constantly. Platforms adjust caption limits, add formats, retire features and update advertising policies, often with little notice. AI models learn from data collected up to a point in time, so they can confidently repeat a rule that changed months ago. Some tools search the web, which helps, but they can still pick up old blog posts or forum guesses. The Academy's guide to how to keep up with social media changes covers the wider habit.
Why AI gets social media facts wrong
| Cause | What happens | Example |
|---|---|---|
| Training cut-off | The model does not know about recent changes | An old maximum video length is quoted |
| Outdated sources | Web search finds old articles ranking well | A retired feature is described as current |
| Blending | Rules from two platforms are mixed together | One platform's ad limit applied to another |
| Confident guessing | The model fills gaps with plausible numbers | An invented "ideal posting frequency" |
| Regional differences | US rules given where UK rules apply | Wrong advertising or data protection guidance |
None of these are rare. Anyone who has asked a chatbot for an image size and then had the platform crop the upload has met the problem.
What AI is good at, and what it is not
| Task | Reliability | Why |
|---|---|---|
| Brainstorming post ideas | High | No single right answer, so errors cost little |
| Rewriting your own copy | High | You supply the facts, the tool supplies phrasing |
| Summarising a document you provide | Medium | Can merge or drop details, so check against the source |
| Quoting platform specs | Low | Figures change often and models lag behind |
| Explaining UK advertising law | Low | Rules are detailed, regional and updated by regulators |
| Predicting algorithm behaviour | Low | Platforms rarely publish exact rules, so answers are guesses |
The pattern is simple. AI is strongest when you give it the facts and weakest when it has to supply them.
Which facts are riskiest
Some answers matter more than others if they are wrong:
- Specs and limits: caption lengths, video durations, aspect ratios and file sizes. A wrong figure wastes production time.
- Eligibility rules: thresholds for monetisation, live streaming or shops. For example, TikTok's requirements for going live are 1,000 followers, 18+ (16+ to watch)✓, and those requirements have changed before.
- Advertising and legal rules: disclosure, competitions, data protection and regulated products. Wrong answers here carry real risk.
- Algorithm claims: AI tools often repeat myths about hashtags, posting times or reach penalties.
Lower-risk uses include brainstorming ideas, drafting captions and summarising your own material.
How to check an AI answer
- Ask the tool for its source, then open that source and check the date.
- Confirm the figure in the platform's own help centre or business documentation.
- For UK rules, check the regulator's guidance, such as the ASA, CMA or ICO.
- Compare with a verified reference that records when each fact was last checked.
- Test it yourself when you can, such as uploading a draft to see how it crops.
How the Academy keeps facts current
The Social Media Academy checks platform facts weekly against official sources. Every change is logged on a public changelog, so you can see what moved and when. You can sign up for free alerts when a fact you rely on changes, and the weekly briefing summarises platform updates, with headlines free and the full briefing in Pro. Lessons pull figures from the same verified source, so a lesson about Instagram captions shows the current limit: 2,200 characters✓. The how we verify page explains the process.
Using AI safely for social media work
AI is still valuable when used with care:
- Use it to draft, then check facts separately.
- Paste verified figures into your prompt rather than asking the tool to recall them.
- Keep a short internal sheet of current specs, updated from a trusted source.
- Ask the tool to flag any statement it is unsure about.
- Never publish legal or regulatory guidance from AI without checking the original rules.
The Academy's guide to what not to automate on social media lists tasks that need a person.
A worked example
Suppose you ask an AI tool how long a TikTok video can be. It might give a figure that was correct a year or two ago. The current answer is 10 minutes (60 minutes for some accounts); sweet spot 21–34 seconds✓. If you planned a series around the old figure, you would either cut content you had filmed or miss a format that now suits you. The same applies to algorithm advice: AI tools often say a platform favours one signal above all others, when official guidance describes several. For example, Instagram's key ranking signals include watch time and completion, likes, shares (sends per reach), saves, relationship history✓. Checking takes a minute and avoids wasted production time.
Spotting a hallucinated answer
Watch for warning signs: very precise figures with no source, references to features you cannot find in the app, links that do not open, and confident statements about how an algorithm "always" behaves. Algorithm claims deserve particular caution. The Academy's guide to social media algorithm myths covers the common ones.
Building a verification habit in a team
If several people use AI tools, agree a simple rule: any figure, rule or legal statement in a post, report or client document must link to a checked source. Add a line to your approval checklist. The Academy's guide to an AI policy for marketing teams shows how to write this down.
The bottom line
AI saves time on ideas and drafts, but it is not a reliable reference for social media facts. Check specs, eligibility rules and regulations against official sources or a verified reference each time, especially before you spend money producing content or run a campaign.
