How the algorithm fit score is calculated

The score is a measurement, not a guess

Every generation is scored 0–100. This score doesn't tell you "how many likes you'll get" — it measures from the text the STRUCTURAL features the X algorithm rewards. The math is transparent: it starts from a base of 50 points, and every measured feature adds or subtracts points. In the breakdown next to the score, you see how many points each item contributes.

You can try the same engine without signing up: the Tweet Scorer (/en/tools/tweet-score) scores any tweet with these exact rules. Scoring runs entirely in your browser — no AI call, and your text is never sent to us.

Measurements that add points

• +strong opening (hook): a curiosity/claim/statistic signal in the first line — graded by signal count and word range
• +ending with a question: an invitation to reply — a reply carries roughly 10× the weight of a like (the question must grow out of the content; a generic question tacked onto the end cancels the reward and gets penalized)
• +conversation starter: questions that invite personal experience — stronger than a plain question
• +list/framework structure: share-worthy content — the kind of thing someone forwards to their team; graded by list depth
• +absence of links: an external link lowers algorithmic reach
• +few hashtags: 0 is ideal, 3 and above is penalized
• +fit to format length: the middle of the target band is full points, the edges less
• +specificity: including real numbers and names/brands
• +quotable sentence: a slogan-like line of 60–90 characters — being quoted carries roughly 10× the weight of a like, which is why this item was raised in the 2026 calibration

If you're curious about the ranking: in X's open-source ranker the strongest positive signal is a SHARE (~40× a like), then replies and quotes (~10×), while a repost is only 2×. The score is calibrated to that order — which is why the "share-worthy" and "quotable" items weigh the most.

Measurements that subtract points

• +engagement bait patterns ("don't forget to like", "RT if...") — X actively penalizes these
• +excessive emoji (more than 5 — increases by count)
• +3+ hashtags
• +use of broken/invisible characters
• +a generic question tacked onto the end ("Do you agree?", "Thoughts?") — the question reward is cancelled and a penalty is added on top
• +formulaic/canned openings (like "Here's what I learned the hard way:") — the opening bonus is suppressed too
• +tool-scented clichés ("the real issue", "the bottom line", "game-changing") — increases by count

What the thresholds mean

• +75+ — the ideal band: a draft carrying the signals X rewards; a good marker for prioritizing your approval queue
• +50–74 — publishable; you can see the low items in the breakdown and regenerate with one click
• +below 50 — a friendly warning: which signals are missing is written out clearly