Platforms change so often that the changes blur together. Someone reads that TikTok has updated something, feels vaguely behind, and carries on. A clearer habit is to ask what kind of update it is, because the kind tells you what to do.
The five kinds of social media platform updates
| Kind | Example | What it can break |
|---|---|---|
| Specification | A limit, length or dimension changes | Templates, scheduled posts, design files |
| Eligibility | A threshold for monetisation or a feature moves | Plans and forecasts |
| Ranking guidance | A platform restates what it rewards | Content priorities |
| Feature | Something new is added or retired | Workflows that relied on it |
| Policy | Rules on disclosure, AI labels or content | Compliance |
The first two are numbers. They are either right or wrong on a given day, which makes them the easiest to track and the most embarrassing to get wrong.
Specification changes: small, silent and costly
Specifications are the limits and shapes your work has to fit. A caption allowance, a maximum video length, a carousel slide count, an image ratio.
They matter because they are baked into everything downstream. A design template is built to a ratio. A script is written to a length. When Instagram's carousel maximum moved to its present 20 slides✓, every guide stating the earlier figure became wrong overnight, and most still are.
These changes are frequently made without an announcement. A help page is edited and nobody is told.
Eligibility changes: the ones that move money
Eligibility rules decide who can use a feature or earn from a platform. Creators plan around them, so a change redraws plans.
The Academy holds them as dated facts. TikTok's requirement for going live is currently 1,000 followers, 18+ (16+ to watch)✓, and YouTube's UK Partner Programme threshold is 1,000 subscribers + 4,000 public watch hours (or 10M Shorts views in 90 days); fan-funding access from 500 subscribers✓. Each has a verified date beside it wherever it appears, because a threshold quoted without a date is a guess about the present.
Ranking guidance: read the source, not the summary
Platforms periodically describe how their recommendation systems work. These statements are valuable and widely distorted. By the time a platform's careful paragraph has been through three newsletters, it has become a rule nobody wrote.
The discipline is to read the original. Whatever a commentator says about a ranking change, the question is what the platform itself said, in its own words. Where the Academy records ranking signals, it records the documented ones and links the source.
Feature launches: mostly not for you yet
New features generate the most coverage and the least immediate consequence. Many launch in one country, or to a tier of accounts, or as a test that is later withdrawn.
Three states are worth distinguishing:
- Testing. Seen by a small group. Ignore.
- Rolling out. Arriving gradually. Note it, and do not plan around it.
- Available. Documented and live on your account. Now it is real.
The Academy applies the same rule to its own facts. A change described as rolling out, in testing or available to some accounts is not treated as a change that has happened.
Policy changes: the ones with consequences
Policy updates alter what you are allowed to do, and they interact with UK law. Platform rules on branded content sit on top of the ASA's requirements and do not replace them. On whether a platform's own paid partnership label is sufficient, the recorded position is: It may be sufficient if upfront, clear and prominent, but ASA research found it "can help consumers recognise ads, but may not be enough on its own" — add a separate "Ad" label if in any doubt.✓
UK regulation changes as well, on its own timetable. Those rules are tracked by the Academy in the same way as platform facts, as versioned entries with primary sources, and summarised on the UK compliance pages.
How a changelog records an update
A list of current values tells you what is true today. A changelog tells you what changed, which is more useful.
Each entry in the Academy's public changelog shows the platform, the fact, the date, the new value and the value it replaced. The previous value is never deleted. That before-and-after is what lets you work out what a change breaks: if a limit went down, anything built to the old limit needs checking.
Corrections are logged the same way. If a fact turns out to have been wrong, the correction appears with the date it was believed and the date it was fixed, and nothing is quietly edited away.
Who decides that an update is real
Automation is useful for noticing and dangerous for deciding. In the Academy's process, software watches the sources and proposes changes, an adversarial pass tries to disprove each proposal and must quote the source word for word, and the quote is then matched against the source page in code. After all that, a person accepts, edits or rejects it. Nothing automated writes to a fact.
The full method is published on the verification page, including the number of facts tracked and the one least recently checked.
Getting updates without reading the news
Three routes, from least effort to most:
- Change alerts. Pick the specifications you depend on from the changelog and get an email the week one moves. Open to anyone.
- The weekly briefing. Members get a summary of what changed and what it means. The headlines are part of the free tier and the full briefing comes with Pro.
- Lesson updates. When a change touches a module you have finished, the revised lesson is flagged on your updates page.
What to do when an update lands
- Identify the kind.
- Confirm it against the platform's own page.
- List what you have built that depends on the old value.
- Fix those things, starting with anything scheduled.
- Note the date, so you know your own records are current.
