X Twitter Engagement Rate: Influencer Stats & Audit

For marketers, creator managers and researchers who want to check an X (Twitter) account before paying for a post. About $0.02 per account. Checking 20 accounts costs about $0.40, plus Apify's small startup fee. No login, account cookies or outside API key is needed.

Get a verdict, engagement rates, follower counts, a reference grade, medians, posting cadence and an HTML report. This audit is based on the last 4 to 5 posts available to signed-out visitors. X normally returns five timeline entries; a pinned entry is separated, leaving four ordinary posts. A pin never enters the averages. This is a screening sample, not a complete history or a prediction of campaign results.

What you get

One account row starts with the answer: its engagement as a percentage of followers, how that compares with the measured reference for its audience size, its grade and the actual sample count. A free summary row, SUMMARY record and REPORT.html accompany every run.

Field	Meaning
verdict	Plain-English result, reference comparison and sample-size disclosure
status	OK, INSUFFICIENT_DATA, UNAVAILABLE or INPUT_NOTE
username, profileUrl	Canonical account handle and public profile link
name	Display name as published on X
followers, following, posts	Current follower, following and lifetime post counts
verified, blueVerified, verificationType	Public verification information; not a trust assessment
postsAnalyzed	Exact number of usable non-pinned posts in the calculation
sampleDescription	Scope: based on the last 4 to 5 posts, with actual usable count
publicPostsFound, postsRequested	Parsed posts and the requested maximum
pinnedPostsExcluded, pinnedPosts	Number of excluded pins and their separate post records
engagementRateByFollowersPct	Mean likes + replies + reposts per analyzed post, divided by current followers, times 100
engagementRateByViewsPct	Total interactions divided by total views, using only analyzed posts with positive public views, times 100
postsWithViews	Number of posts supporting the view-based rate
averageLikes, averageReplies, averageReposts	Average interaction counts per analyzed post
medianLikes, medianReplies, medianReposts	Median interaction counts, reducing the effect of one unusually successful post
averageViews, medianViews	Average and median among available view counts; null when unavailable
averageQuotes, medianQuotes	Separate quote-post counts where available; excluded from the engagement formula
replySharePct	Received replies as a percentage of likes + replies + reposts
postingRatePerWeek	Estimated cadence from intervals between the analyzed original posts
sampleWindowDays, daysSinceLastPost	Time covered by the sample and age of its newest post
oldestPostAt, newestPostAt, checkedAt	UTC timestamps for the sample and measurement
sizeTier, normForSizePct	Audience-size band and median engagement rate in the fixed reference sample
grade, referenceRatio	A-D comparison with that reference, and rate divided by reference
referenceAccounts, referenceDate, referenceMethod	Cohort size, date and limitations of the comparison
analyzedPosts	Every post included, with URL, timestamp, text and individual counts
topPosts	Up to three analyzed posts ranked by likes + replies + reposts
notes	Missing views, stale posts, capped requests and other limitations
charged	Whether this row represents a charged account audit; summary and note rows are false

When to use it

Shortlist crypto creators, compare fitness accounts, or check tech creators before buying a post. Agencies get a consistent comparison table; researchers get dated counts and source links.

How to use it

1. Enter handles in X usernames, such as NASA or @levelsio. Alternatively, paste X or Twitter profile URLs. You can use both lists.
2. Leave Posts to analyze at 5 and review the run's maximum charge. The prefilled two-account example costs about $0.04 plus startup.
3. Start the Actor. Open the overview dataset view for verdicts, the metrics view for comparisons, or the linked HTML report. Export the dataset as JSON, CSV or Excel from Apify.

Handles and URLs are deduplicated case-insensitively. Post URLs generate free input notes. Empty input does not scrape any accounts.

Inputs

Input	Default	Examples and behavior
usernames	None at API level; console prefill contains two accounts	["NASA", "levelsio"]; with or without @
urls	Empty	["https://x.com/NASA", "https://twitter.com/levelsio"]
postsToAnalyze	5	Requested maximum. At most five public entries are available; a pin reduces the usable count. Values above five are capped with a note.
saveRawData	false	Save public source pages in the run's RAW key-value record for verification. Adds storage.

The API also accepts startUrls entries such as {"url":"https://x.com/NASA"} for integrations that already use that field. All lists are combined and deduplicated. A run accepts at most 1,000 input entries; split larger work into batches. There is no special free-plan cap.

{
  "usernames": ["NASA", "levelsio"],
  "postsToAnalyze": 5,
  "saveRawData": false
}

Requests above five cannot unlock history. Invalid counts use the default with a free note. Fewer than three usable posts or no usable follower denominator produces a free Insufficient data row.

How the calculations work

Follower engagement = mean(likes + received replies + reposts) / current followers * 100. The denominator is today's follower count. Quote counts are separate and excluded. Edited-post counts include prior revision interactions.

View engagement = sum(interactions) / sum(views) * 100, restricted to the same posts with positive public views. Missing views remain null; those posts can still contribute to follower engagement. postsWithViews shows the denominator coverage.

Posting cadence = (posts analyzed - 1) / days between oldest and newest * 7. It covers observed original-post intervals, excluding the account's reply and repost activity. replySharePct measures received replies as a share of interactions, not how often the account replies.

Pins are identified from X's structured pin instruction and reported separately. They never enter averages, medians, cadence or grades. Quoted posts and other authors' posts are excluded. Missing counts are not zero.

Reference grades by audience size

The reference is our own dated convenience sample, not a population-wide X benchmark. On 14 October 2026, the calibration tested 142 public handles across technology, developers, fitness, crypto, media, sports, organizations and consumer brands. There were 126 usable account audits. Removing accounts whose newest sampled post was over 90 days old left 94 accounts for the reference.

Followers	Reference accounts	Median account engagement
Under 100,000	22	0.191482%
100,000 to under 1 million	36	0.070086%
1 million to under 10 million	20	0.033506%
10 million or more	16	0.005433%

Medians are across account-level rates, without weighting larger accounts more heavily. This selected sample mixes niches and account types. Same-size fitness creators and news publishers may differ greatly. The reference is fixed, not recomputed from your run.

Grades compare your rate with the reference median: A is at least 2x; B is 1x to under 2x; C is 0.5x to under 1x; D is below 0.5x. They are comparison bands, not population percentiles. Cohort size and date accompany each grade.

Pricing and spending limits

The list price is $0.02 per useful account audit. Bronze is $0.019, Silver $0.018, and Gold and higher tiers $0.016. Apify's startup fee remains $0.00005 per GB, with a minimum of one startup event. The default memory is below 1 GB, so the minimum startup fee applies. Platform usage is included in this Actor's pay-per-event price; no external service key is needed.

Useful accounts	List-price account charges	Bronze account charges
2	$0.04	$0.038
20	$0.40	$0.38
100	$2.00	$1.90
1,000	$20.00	$19.00

Add the startup fee to each example. Failed lookups, insufficient samples, invalid inputs, duplicates, retries and summary rows do not incur an account-audit charge. A run containing only free notes can still incur Apify's startup fee.

The Actor checks the spending limit before another fetch, charges only after usable data arrives, and handles the row write and charge together. The summary reports a spending-limit stop; increase your limit or split the remaining accounts into another run.

Real output example

From a private cloud verification on 14 October 2026. Counts change over time.

{
  "verdict": "Engagement 0.0095% of followers per post, at or above reference for 10M+ followers (B); based on the last 4 to 5 posts (5 analyzed).",
  "username": "NASA",
  "followers": 92328883,
  "postsAnalyzed": 5,
  "pinnedPostsExcluded": 0,
  "engagementRateByFollowersPct": 0.009451,
  "engagementRateByViewsPct": 0.591187,
  "grade": "B",
  "normForSizePct": 0.005433
}

API examples

Use your own Apify token. For large lists, use asynchronous runs. Filter type == "ACCOUNT" and status == "OK" to exclude notes and summaries.

curl -X POST \
  'https://api.apify.com/v2/acts/vant47~twitter-engagement-rate/run-sync-get-dataset-items?timeout=120' \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H 'Content-Type: application/json' \
  -d '{"usernames":["NASA","levelsio"],"postsToAnalyze":5}'

import os
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])
run = client.actor("vant47/twitter-engagement-rate").call(
    run_input={"usernames": ["NASA", "levelsio"], "postsToAnalyze": 5}
)
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    if row.get("status") == "OK":
        print(row["username"], row["verdict"])

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('vant47/twitter-engagement-rate').call({
  usernames: ['NASA', 'levelsio'], postsToAnalyze: 5
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.filter(row => row.status === 'OK'));

Use with AI agents (MCP)

Connect to https://mcp.apify.com. Ask your agent to search Twitter engagement rate, inspect the schema, and pass usernames. Example: “Compare these X tech creators before I buy a post. Include followers, views, sample size and limitations.”

Agents should preserve the sample caveat, check post ages, explain the selected reference and inspect spendingLimitReached. Grades are not fake-follower findings or earnings forecasts.

Saved tasks and integrations

Five saved tasks cover crypto, fitness, tech creators, brands and space, each with two accounts. Duplicate a task and replace its handles. Exports work with spreadsheets; Apify supports schedules, Make, Zapier, n8n and webhooks. No schedule is enabled automatically.

For repeated measurements, save timestamps, post IDs and counts. New posts change the sample, so rate changes need context. The Actor does not retain history or calculate follower growth.

Limits and FAQ

Why four or five posts? X shows five signed-out timeline entries. A pin leaves four ordinary posts; it is reported separately. Three usable posts can qualify, with the actual count disclosed. Fewer than three is free. Requesting more cannot unlock history.

Protected, deleted or suspended accounts? Unavailable accounts receive free notes. The calibration produced 16 unavailable or insufficient cases without account charges.

Are counts exact? Structured public numbers are used, not rounded labels. They reflect the response at measurement time, subject to X's caching. Missing values stay null. Counts may change before you visit a post.

Does an A guarantee performance? No. Viral content, promotions, niche differences and immature engagement can distort a short sample. Grades use the disclosed reference, not a representative X-wide population.

Bookmarks, clicks or impressions? Not included. Public views are not unique viewers or private analytics impressions. Quote counts are separate from the headline formula.

Old or very recent posts? Notes flag sampled posts under 24 hours old and accounts whose newest post is over 30 days old. The data is never presented as a complete current-month history.

Can I verify it? Every analyzed post includes a source URL and counts. Enable saveRawData for public source HTML in RAW. The report links to sampled posts.

What if X changes? Datacenter requests are retried, then residential is attempted. Unusable data receives a free note. No login, borrowed account cookies, CAPTCHA solving or paid data API is used.

Affiliated with X? No. This is an independent tool. Review X's terms and your intended use before collecting or distributing data.