I downloaded every original post from @antonships and audited 221 of them. The archive contained 78,838 impressions, 1,684 replies, 53 bookmarks, 85 long-form notes, and no native X Articles at the time of export.
Those totals describe the archive window. They do not prove revenue, audience quality, or what a new account should expect. Their value is that they replaced my memory with a complete set of posts I could label and compare.
Export your own posts, keep the original URL and date, and record the visible metrics from one consistent snapshot. If the export omits a metric, leave it blank rather than reconstructing it from vibes.
Complete post setOne measurement windowUnknown stays unknown
02
Give every post one primary job
The biggest mistake in my first audit was treating reach as a verdict on the whole post. A connection question and a product lesson are not trying to do the same thing.
Discovery posts create replies and expose the account to more people. Authority posts carry evidence, a useful model, or a lesson someone might remember. Conversion posts connect a demonstrated problem to a product, launch, waitlist, or trial. A post can help more than one job, but choose the primary one so the comparison remains useful.
DiscoveryAuthorityConversion
03
Label the mechanism, not just the topic
Two posts about founder marketing can behave differently because one is a broad question and the other contains a screenshot, a failure, and a conclusion.
Record the opening type, format, evidence type, audience, and call to action. Useful labels include constraint question, specific failure, numbered lesson, product receipt, strong opinion, launch, and direct ask. Keep the vocabulary small enough that you can apply it consistently.
OpeningFormatEvidenceCTA
04
Compare within a job before comparing across jobs
Rank discovery posts against other discovery posts. Rank evidence-led authority posts against comparable authority posts. Only then look at how the mix supports the account as a whole.
My high-reply posts made me visible, but the posts that explained a real Pocket Orbit or Sendezeit problem did more to make the account credible. The conclusion was not that conversation posts were bad. It was that I had been asking one metric to grade three jobs.
Like-for-like comparisonMedian before outlierRead the replies
05
Turn the audit into next-week decisions
An audit should end with a small publishing decision, not a museum of old screenshots. Keep one discovery format, one authority mechanism, and one conversion bridge worth testing again.
Also write down what to stop. In my case, that included repeating broad founder questions too often and teaching a writing system from every post equally. Some posts are useful for distribution and still poor examples for voice.
Download the CSV template, add your archive, and review a manageable batch before trying to score years of history in one evening.
One format to keepOne mechanism to testOne habit to stop
Continue through the topic
Product, tool, and evidence stay connected.
People also ask
Direct answers, not hidden objections.
What should a social media content audit include?
Include the post, date, URL, format, topic, primary job, evidence type, CTA, comparable metrics, qualitative replies, and the next decision.
How many X posts should I audit?
Start with the most recent 30 to 50 if a full archive is too large. Use a complete time window rather than cherry-picking favourites.
Are impressions enough to judge an X post?
No. Impressions help describe discovery, but authority and conversion need other evidence such as saves, useful replies, profile actions, clicks, signups, or sales where attribution exists.
Can AI choose my best writing samples?
It can help sort candidates, but the final Keep or Pass decision should remain yours because high reach and representative voice are not the same thing.
Three-day trial
Turn the framework into a real publishing system.
Choose the plan before creating the account. The exact charge date appears before checkout.
