Every agency has a version of this phone call. Three weeks into a campaign, the client forwards a screenshot: one of your creators, two years ago, joking about something the client's legal team does not find funny. Or a journalist points out that the "community" cheering a sponsored post is mostly emoji strings from accounts with no faces. Nobody on your team did anything wrong, exactly. Somebody ran out of time to look. That is what an influencer vetting failure looks like from the client's side: not a process gap but a judgment question, aimed at you.
Influencer vetting is the checking stage between finding a creator and contacting one: is the audience real, is the engagement genuine, is the content history safe for this particular client, and has the creator already worked with a competitor? It sits between discovery (building the longlist) and monitoring (watching what happens after signature). Of the six steps behind the roughly ten hours teams tell us a single creator takes (the manual work we walked through in The 10-hours-per-creator problem), vetting is the one teams most often shortcut when a deadline bites. Understandably so: it is slow, screen-by-screen work that produces nothing visible. Sourcing produces a list. Outreach produces replies. Vetting, done properly, mostly produces the absence of a disaster, which is exactly why it gets squeezed, and exactly how campaigns end up in client escalations.
What influencer vetting has to cover
Strip away the tool-vendor framing and vetting is four questions per creator.
Is the audience real? Followers can be bought, and so can likes, comments and story views. Check that the audience plausibly matches the content: the right countries and languages for your client's market, a growth curve that climbs steadily rather than in overnight cliffs, and engagement in sane proportion to follower count. People occasionally ask who the "fakest influencer" is, as if fraud had a face. It does not: fake reach is a property of accounts, not a list of villains, which is why you check every account, not just the suspicious-looking ones.
Is the engagement genuine? This gets its own section below, because it is where most vetting quietly goes wrong.
Is the content history safe for this client? Scroll back twelve to eighteen months. Jokes, politics, drinking, old controversies, general tone. Safety is relative: content that is fine for a festival brand can sink a bank. The question is never "is this creator safe?" but "is this creator safe for this client?"
Any past brand conflicts? Previous deals with a competitor, a category conflict the client cares about, a feed so dense with sponsored posts that one more will vanish in it. Disclosure habits belong here too. More on that below.
The engagement-rate trap
When we ask agency teams to walk us through how they vet influencers, the most common shortcut is a single number: average engagement rate (likes and comments as a share of followers) checked against a threshold, box ticked. Two problems.
First, the average lies. One pinned post that went viral can drag a creator's average up several times over while the last twenty posts sat nearly unread. The honest signal is the median across recent posts: the middle value when you line the posts up, which one freak hit cannot move. Take the last fifteen to twenty posts and ask what a typical one does.
Second, single-number thresholds fail in both directions at once. They reject real talent (a niche creator with a small, loyal audience can sit "below threshold" and still outsell an account three times the size) while admitting gamed accounts, because engagement rate is precisely the number that engagement pods (groups of accounts that like and comment on each other's posts to look popular) are built to inflate. No single figure separates the two. Several signals together do: engagement relative to follower size, comment quality (specific, on-topic sentences versus emoji walls and "great pic!"), growth pattern, and posting consistency. Any one of these can be gamed cheaply. Gaming all four at once, convincingly, is hard enough that few bother.
The average engagement rate is the easiest number in a media kit to inflate. The median across recent posts is the honest one.
Vetting changes with follower size
When people ask about "the four types of influencers", they mean follower bands, nothing deeper. The bands most of the industry uses: nano at roughly 1–10k followers, micro at roughly 10–100k, macro above that, and mega (celebrities) at the top.
How to vet influencers depends on the band. At the macro and mega end, the record is long, public and searchable, so brand safety and past partnerships dominate the check. At the nano and micro end, the risk flips to authenticity: less public history, and this is where bought followers and pod activity concentrate. The catch is volume: a strategy built on sixty small creators (often the right strategy, as we argue in The case for nano and micro influencer campaigns) means sixty vetting passes instead of fifteen. That workload, more than performance, is why those strategies get flattened into fewer, bigger names. The ops burden shapes the strategy, backwards.
An influencer vetting checklist you can run tomorrow
One pass per creator, before outreach starts. Timebox it (a fixed slot in the diary), because unbounded vetting either eats the week or gets skipped.
- Audience fit: country and language split match the client's market
- Growth curve: steady climb; investigate any vertical jump
- Median engagement: across the last fifteen to twenty posts, not the average
- Proportion: engagement plausible for the follower count
- Comments: read twenty; real sentences or emoji walls?
- Content history: twelve to eighteen months back, judged against this client
- Past deals: competitor work, category conflicts, sponsored-post density
- Disclosure habits: are old ads clearly labelled?
- The log: pass or fail, why, and who signed off
That last line is the one almost nobody keeps, and it is the cheapest insurance on the list. When a creator does blow up mid-campaign (run enough campaigns and eventually one will), the difference between an awkward call and a lost account is showing what you checked, when, and why the creator passed at the time. One line per creator in a shared sheet is enough. Vetting you cannot evidence might as well not have happened.
Disclosure is a vetting question, not a legal footnote
Disclosure (the legal requirement that creators clearly label paid partnerships as advertising) usually gets handled in the brief and the contract. Check it at vetting instead, because a creator's back catalogue tells you the truth: do they label ads clearly and up front, or bury #ad twenty hashtags deep? The exact rules vary by market (European regulators require paid partnerships to be clearly and visibly labelled, and the US FTC enforces the same idea), but the predictive signal is the same everywhere. A creator with a habit of undisclosed partnerships will bring that habit to your campaign, and when a regulator or a journalist notices, it is the brand and the agency doing the explaining, not just the creator.
Where influencer vetting tools help, and where the hours stay
Influencer vetting tools (Modash, Upfluence and the rest) are genuinely useful for the first pass: fake-follower estimates, audience geography, growth charts on demand. Use them. But two things stay manual: the judgment (is this creator right for this client?) and, surprisingly, much of the checking itself: reading comments, scrolling content history, cross-referencing past deals against your client roster, logging the decision. Screen by screen, creator by creator.
That second layer is rule-following work, and it automates well: pull the recent posts, compute the median, flag outlier spikes, surface competitor mentions, write the log entry. Then hand the borderline cases to a human for the call. The judgment does not move; the drudgework around it does. It is the same pattern we describe in Influencer outreach that doesn't eat your week: automate the chase-and-check layer, keep the decisions. Automated step by step, that layer is how a team that caps at twenty creators today could run a hundred or more, with the same people making the calls. And the bar should be high: we treat a pilot as failed unless manual time on the automated step drops by at least 70%.
Start with your last shortlist
Here is the practical version. Take five creators from your last completed campaign and re-vet them for an hour: median engagement instead of average, twenty comments read, growth curve eyeballed, past deals checked against the client's competitors. If all five come back clean, your process is better than most. If one or two come back wobbly, you have learned something more useful than any benchmark: where your own vetting bends under pressure. Fix that first (tighten the checklist, timebox the pass, keep the log), and the next difficult client call becomes one you answer with a record instead of a shrug.