How Advertisers Detect Inflated Influencer Numbers

Bought views are cheap. A few dollars per thousand, delivered overnight, indistinguishable from real ones if all you look at is the view counter. Which is why smart advertisers stopped looking at the view counter alone years ago. Real audiences behave in ratios, and bot traffic almost never fakes all the ratios at once. If you know what the ratios should look like, inflated numbers stop being invisible and start being obvious.

This is a practical guide to running those checks yourself with live data. Every number you need is one $0.01 API call away, so vetting a creator costs pennies, not a subscription to an influencer platform. We run these exact screens on our own money, which we will get to.

The core insight: fraud is expensive in every dimension except one

A vendor selling views sells views. Some also sell likes. Very few sell views, likes, comments, shares, and believable follower growth in a coordinated, proportional package, because at that point the fraud costs more than the real thing. So inflated accounts are almost always lopsided: one metric towers, the others lag. Your job is to measure the lopsidedness.

The four screens

1. Likes per view

Healthy short form video runs a like rate of roughly 3 to 7 percent of views, with exceptional content pushing past 10. When you see 2 million views carrying 9,000 likes, a 0.45 percent rate, something bought the gap. The inverse pattern exists too: purchased likes on modest views produce absurd 25 percent like rates. Both tails are flags.

2. Comments per view

Comments are the hardest metric to fake convincingly because they require content, so this ratio breaks most fraud. Real viral posts typically land between 0.05 and 0.5 percent comments per view. A million views with 40 comments is a screaming outlier. If you spot-read the comments and they are generic emoji strings from accounts with no posts, you are done investigating.

3. Views per follower

Pull the creator's follower count and divide. Organic distribution on TikTok and reels means good creators regularly exceed 1x their audience, and breakouts hit 10x or more. But a mid-size account whose every post does 30x followers, with weak likes and comments underneath, is showing you paid distribution, not talent. Consistency is the tell: real virality is spiky, bought reach is smooth.

4. Trajectory shape

Real view curves decay: fast growth in the first hours, then a long flattening tail. Bot deliveries arrive in batches, so the curve shows stair steps, or a dead post that suddenly gains 400,000 views in week three. You can only see this by reading the same post repeatedly, which is exactly what per-request pricing makes trivial.

Running the screens in code

The x402 Social Stats API returns every input these checks need. One call per post, one call for the profile, $0.01 each via the x402 protocol, no account or key:

import { wrapFetchWithPaymentFromConfig } from "@x402/fetch";
import { ExactEvmScheme } from "@x402/evm";
import { privateKeyToAccount } from "viem/accounts";

const account = privateKeyToAccount(process.env.AGENT_PRIVATE_KEY);
const payFetch = wrapFetchWithPaymentFromConfig(fetch, {
  schemes: [{ network: "eip155:8453", client: new ExactEvmScheme(account) }],
});

async function screen(postUrl) {
  const r = await payFetch(
    "https://x402.findclout.com/api/v1/post?url=" + encodeURIComponent(postUrl)
  );
  const s = await r.json();
  if (!s.success) return { postUrl, verdict: "unreadable" };

  const likeRate = s.likes / s.views;
  const commentRate = s.comments / s.views;
  const flags = [];
  if (likeRate < 0.01) flags.push("likes lag views");
  if (commentRate < 0.0005) flags.push("comments lag views");
  if (s.followers && s.views / s.followers > 30 && likeRate < 0.02)
    flags.push("reach outruns audience with weak engagement");
  return { postUrl, views: s.views, likeRate, commentRate, flags };
}

The universal /api/v1/post endpoint auto-detects Instagram, TikTok, and X from the URL, and on TikTok and X it includes the author's follower count in the same response. For Instagram, grab followers with one extra call to the profile endpoint. Treat the thresholds above as starting points and tune per platform: baseline ratios differ, and photos on Instagram return views: null, so screen those on likes-to-followers instead.

For trajectory checks, run the same screen on a schedule and diff the readings. A creator vet across their last 12 posts plus profile is 13 cents. Re-reading a sponsored post daily through a 30 day flight adds 30 cents. The complete scheduled-verification loop, with delivered CPM math, is written up in how agents verify influencer campaign results automatically.

When to run which check

Before signing: screen the creator's recent posts for ratio health and consistency. Decline quietly if two or more screens flag.
During the flight: poll deliverables daily. Stair-step growth mid-flight is your cue to ask questions while leverage still exists.
At settlement: verify final numbers yourself rather than accepting screenshots. Screenshots are free to edit; live API reads are not.

Billing works in your favor while you automate this: malformed URLs return HTTP 400 and 4xx responses are never charged, while a paid {"success": false, "error": "not_found"} on a post that was supposedly delivered is itself a finding worth having.

Why we are opinionated about this

FindClout is on both sides of this trade every day. We operate the biggest independently owned American meme page network, spanning sports, finance, entrepreneurship, movies, TV, and music, advertisers buy native placements inside our viral clips, and creators on our network get paid based on verified views. Those payouts only work if the views are real, so ratio screens like these run against our own campaigns before a dollar moves, using the same endpoints documented on this site. Inflated numbers are not an abstract industry problem to us; they are a line item we refuse to pay. The longer story is in why FindClout built this API.

Vet before you pay. The quickstart gets your first screen running in minutes, and the whole vetting habit costs less than one fake thousand views would have.

How advertisers detect inflated influencer numbers