How AI scoring works

Every person the Leads module captures from your posts is scored against your ICPs. Each scoring produces a fit level and the reasoning behind it, so you can trust the result or overrule it.

Scoring is the one thing MagicPost meters: 1 credit = 1 lead scored against 1 ICP. A paid plan (Creator and above) comes with 100 credits per month; the free trial gives you 10 credits, valid 3 days. Everything around the scoring, including the pre-qualification pass below, is free.

The four fit levels

• Excellent. Strong match. Prioritise them.

• Good. Worth a conversation.

• Medium. Weak match. Useful for nurture, not for active outreach.

• Low. Not a match on this ICP.

A Low lead is not an error. It passed pre-qualification (nothing in its profile ruled it out up front), and the AI then read the full profile and judged it far from your target. It stays in your list, and you decide whether it is still worth a message.

The two steps

1. Pre-qualification: free, no credit spent.

Each profile is checked against your ICP's keywords and filters. There are three outcomes.

• Ruled out. An exclusion keyword hits, or the profile fails a hard filter: country include-list, an exclusion on country, industry, company size or city, follower range, or LinkedIn profile type. The lead is marked Disqualified and never sent to the scoring AI. No credit.

• Clear match. An indispensable keyword hits, or two secondary ones do, and no filter objects. The profile goes straight to scoring.

• Borderline. Not enough keyword hits, or the profile misses your industry, company size or city include-list. This does not disqualify anyone: those LinkedIn fields are free text and too unreliable to reject on. A lightweight AI check reads the profile and decides whether it is worth a full scoring. That check is free too, and it either passes the profile on or marks it Disqualified.

Two more things worth knowing. A filter is skipped when the profile has nothing in that field (a profile with no country listed is not rejected by a country filter). And no credit is ever spent at this step, whatever the outcome.

2. AI scoring: 1 credit per lead, per ICP.

The profiles that got through are read by the AI and compared to your ICP. This is what produces the fit level and the explanation.

What the AI reads

Only the person's LinkedIn profile:

• Full name, headline, current role

• Company, industry, company size

• Country

• Their About section

• Their company's description

It does not read their own posts, their activity elsewhere on LinkedIn, or the comment they left under your post. The comment is captured and displayed on the lead's record, but it plays no part in the score. Fit is about who the person is, not what they wrote.

What you get on each scored lead

Open a lead and go to its qualification tab. For each ICP you see:

• The fit level: Excellent, Good, Medium or Low.

• A key quote: a short, literal excerpt from the profile that carried the decision.

• The reasoning: a few short bullets explaining the verdict.

• The signals that pushed the score up or down.

• The pre-qualification detail: whether the profile matched or was disqualified, which keywords hit, and which exclusion or filter rejected it.

When the AI gets it wrong

The score is a recommendation, not a verdict. The reasoning shows you exactly why it landed where it did, so you can spot a miss: an ambiguous job title, a profile with nothing written on it.

You can override the verdict yourself: select leads on a post and mark them Match, Not a match or Uncertain. Your verdict wins over the AI's, and it is free. Adding a lead to an ICP's list marks it a match; removing it marks it a non-match.

Repeated misses usually mean the ICP itself needs work (see How to edit an existing ICP).

What costs credits

A credit is the unit MagicPost meters AI scoring with, and understanding and buying Leads credits covers your quota and your top-ups.

| Action | Cost |

|---|---|

| Pre-qualification, including the borderline check | Free |

| AI scoring of a new lead, per ICP | 1 credit |

| Re-scoring a lead after you regenerate an ICP | 1 credit |

| Your own manual verdict | Free |

| Capturing the engagers on a post | Free |

| Editing, archiving or deleting an ICP | Free |

FAQ

The same person matches two of my ICPs. Does that cost two credits?

Yes. A credit buys one analysis against one ICP. Two ICPs means two analyses, each with its own level and its own explanation, so two credits. This is why the detection screen lets you choose which ICPs to run against.

Why is a lead "Low" if it was pre-qualified?

Pre-qualification is a cheap keyword and filter pass: it only rules out the obvious misses. The AI then reads the whole profile and can still conclude the person is far from your target. That is the system working, not failing.

Does a manual verdict re-run the AI?

No. It records your judgement alongside the AI's score without replacing it, and it costs nothing.

Need help?

Reach out via the in-app chat, or book a 30-minute onboarding call: https://cal.com/magicpost-team/contact-sales-team