MagicPost scores every captured engager against your ICPs using AI. Each scoring produces a level (A, B, C, or D), a score out of 100, plus reasoning so you can trust and audit the result.

The score levels

• A: Strong match. This person fits your ICP very well, prioritise them.

• B: Good match. Worth a conversation.

• C: Weak match. Might be useful for nurture, not for active outreach.

• D: Not a match. Filter them out.

A numeric score (out of 100) is shown next to the level for finer comparison between two leads at the same level.

The two-step pipeline

1. Pre-qualification (free): each profile is checked against your ICP's keywords and default filters. Profiles that obviously do not fit are marked excluded and skipped. No credit charged.

2. AI scoring: the remaining profiles are scored by the AI against your ICP. This is where the level, score, intent, signals, key quote, and reasoning come from. Uses credits from your Leads quota.

What you see for each scored lead

• Level + score: A/B/C/D plus a 0-100 score

• Intent: what the lead seems to want, based on their profile and engagement

• Signals: concrete data points that pushed the score up or down

• Key quote: a short verbatim from the profile or their post engagement

• Reasoning: the AI's full explanation

Hover the qualification badge in the leads table to open a details popover with all of the above.

Manual override

If you disagree with the AI's score, click the qualification cell and pick your own verdict: Match, Not a match, or Uncertain. Your verdict takes priority over the AI's and is visually distinct in the leads table. Pick Clear the verdict in the same popover to revert to the AI's score.

What costs credits, what does not

• Pre-qualification: free

• AI scoring of a new lead: uses credits

• Re-scoring after you edit the ICP: uses credits

• Manual verdict, clearing a verdict: free

What if the AI gets it wrong?

The score is a recommendation, not a verdict. The reasoning panel shows exactly why the AI scored as it did, so you can spot misses (missing context, ambiguous job title) and override. Recurring misses usually mean the ICP prompt or positive examples need refining, see How to edit an existing ICP.