Roast & Optimize Your X Profile
Enter a handle and the tool scores the public profile out of 10 against measurable checks - photo, bio, link, follow ratio, posting cadence and what a typical post earns - then lists the fixes, biggest first. Every number is measured, not guessed, and the same profile gets the same score.
Checking your browser...
- Public data only
- No password needed
- Nothing is posted to the account
How it works
- 1
Enter a handle
Any public X username. We read the bio, the account stats and the recent posts, reposts excluded. No login, nothing posted.
- 2
Eight measured checks
Photo, header, bio, link, follow ratio, posting cadence, engagement and media add up to a score out of 10, every check with what it found.
- 3
Fix the biggest first
The fixes come ordered by the points they are worth. With an AI model configured, a roast and three suggested posts come on top.
What the score checks
Profile photo
Header image
Bio
Link
Follow ratio
Posting cadence
Engagement on original posts
Images and video
What the score is built from
The tool reads a fixed set of public facts: the handle, display name and bio, the follower, following and total-post counts, the profile photo and header image, the website and location fields, and the account's recent posts with reposts filtered out. Eight checks score them: a custom photo, a header image, the bio (graded by the same rules as our bio optimizer), a link, the follow ratio, posting cadence, what a typical original post earns, and how often posts carry images or video. The points add up to a score out of 10, and the fixes are listed in the order of points they cost.
Every threshold is stated. The follow ratio is only judged once an account follows more than 500 others. A last post more than 60 days old reads as dormant, and 3 or more posts a week counts as a steady cadence. Engagement is the median interactions per original post as a share of followers, judged in absolute numbers below 100 followers where a percentage is noise. A check whose input we could not read, such as cadence when only top posts were available, is marked not measured and left out of the score rather than failed. The same profile gets the same score every time.
When a language model is configured, it adds a comedy roast and 3 suggested posts under X's 280-character limit, written from up to 20 recent posts, and the result then shows the model's own score and names the model. Without one, you get the rule-based review, labelled as such, and no invented jokes or stock posts.
Context for the follower count: across the 459,415-account sample PlayerSells measured from its 22,837,969-account X index on 5 October 2026, the median account has 490 followers, the 90th percentile has 6,183 and the 99th has 76,100. 10,000 followers, the number most growth advice treats as a starting line, is larger than 93.2% of that index. A 490-follower account is the middle of X, not a failure, which is why no check here scores the follower count itself.
The eight checks, their weights and their thresholds are this tool's own design. Profile and post data are read from public X data. The follower percentiles are our own measurement, taken on 5 October 2026 over a 459,415-account sample of the 22,837,969 X accounts we index, and describe that index rather than every account on X.
What comes back
| Section | What it is |
|---|---|
| Score out of 10 | The points from eight measured checks, banded on screen as Strong, Solid, Needs work or Weak. The same profile gets the same score every time. |
| Measured checks | What every check found and the points it earned. An input we could not read is marked not measured and left out of the score. |
| Fixes, biggest first | The most useful part of the result: every line names one specific thing about the profile you can change today, in the order of points it costs. |
| The roast and Posts written for this account | Only when an AI model is configured: a comedy roast written from the account’s own posts, and 3 posts under X’s 280-character limit. Without a model these sections are not shown at all, rather than filled with stock text. |
What you get without AI
When no language model is configured, or the one that is fails, the tool does not error out. It returns the profile review: the score out of 10, every measured check with what it found, and the fixes in order of the points they cost. The page labels it a rule-based review, because that is what it is.
What it will not do is dress that review up as a model’s writing. There is no stock roast, no templated joke about your follower ratio and no canned “posts you should write”: those sections only appear when a model actually wrote them, and the result then names the model. A fabricated roast would read like an opinion about your profile that nobody formed, and the measured findings are the part worth acting on anyway.
Frequently asked questions
What is the Profile Roaster?
The Profile Roaster reads a public X (Twitter) profile - bio, stats and recent posts - and scores it out of 10 against measurable checks, then lists the fixes in the order of points they cost. When an AI model is configured it also writes a comedy roast and three suggested posts on top of that.
How is my profile analyzed?
It reads public profile data - bio, display name, follower and following counts, total posts, profile photo, header image, website and location - plus the account's recent posts. Eight checks score it: a custom photo, a header image, the bio (graded by the same rules as our bio optimizer), a link, the follow ratio, how often and how recently it posts, what a typical original post earns relative to followers, and how many posts carry images or video.
Is it really free?
Yes, with no sign-up. Guests get one analysis a day and a free account gets five per tool per day. The review is computed from rules, so it works whether or not an AI model is available.
Does it need AI to work?
No. Without a model you get the profile review: the score out of 10, every measured check and the fixes, clearly labelled as rule-based. When an AI model is configured and reachable, it adds a comedy roast and three suggested posts, and the result says which model wrote them. If the model fails, you get the review rather than an error.
How accurate is the score out of 10?
In the profile review it is arithmetic: eight checks with fixed weights, so the same profile gets the same score and every point lost is shown with its reason. A check whose input we could not read, such as posting cadence when only top posts were available, is marked not measured and left out of the score rather than failed. When a model writes a roast, the number is the model's own judgement and can move between runs.
Can I share my result on X?
Yes. There's a 'Share on X' button right on the results page. The shared post says whether the score came from the rule-based review or from an AI roast.
Will the roast be mean?
The profile review is not a roast at all: it is plain findings and fixes. When a model writes a roast, it is savage but never mean-spirited, and the measured checks and fixes still come with it.
More Free X Account Tools
X Trends
what is trending right now, by country
Algorithm Scorecard
which signals the ranking model rewards or ignores
Tweet Analyzer
why one post travelled and another did not
Best Time to Post
the hours an account's own audience actually shows up
Telegram Best Posting Time
which weekday and hour a channel's posts actually travel furthest
Telegram Post Analyzer
which posts spread on a channel and what they had in common
Telegram Trends
which telegram channels, topics and languages are gaining fastest
Bluesky Best Posting Time
when an account's own posts actually land, as a 7 by 24 heat map
Browse X (Twitter) rankings
the whole tracked market, sorted and filtered
All free tools
every calculator, checker and ranking tool in one place
Buying or selling X (Twitter) accounts?
PlayerSells holds the buyer's payment in escrow until the account has been handed over and checked, so neither side has to trust a stranger.
PlayerSells is not affiliated with X Corp. In the profile review, the score out of 10, the findings and the fixes are computed from measured checks with fixed thresholds. The roast and the suggested posts are generated by a language model only when one is configured, and in that case the score shown is the model’s own judgement and the result names the model. Profile and post data are read from public X data, from a stored snapshot up to 24 hours old where one exists, and the result is cached for 10 minutes. Follower percentiles are our own measurement, taken on 5 October 2026 over a 459,415-account sample of the 22,837,969 X accounts we index.