AI Content Scoring Agent: how content gets scored

Of the four AI Workforce agents, this is the one to turn on first — it's the cheapest per match, and every other agent's output is more useful once content is already scored. This article is the deep dive; the setup order across all four agents is in Configure Your AI Workforce.

What it actually scores

Every piece of content gets four numbers, each out of 100:

• Brand Safety — is this safe to associate your brand with
• Hook — how well the opening seconds or lines earn attention
• Product Fit — how clearly the content ties to a specific product in your catalog
• Content Quality — overall structure and clarity

Why it's the cheapest agent

At 1 credit per match, it's priced below the other three (2 credits each) — reflecting that scoring is a lighter-weight task than matching content to products and creators, or reconciling ad spend. That's also why it's the recommended first agent to turn on: cheap, foundational, and immediately useful even before the others are active.

How scoring actually works: a visual model, not just text

This isn't a text-only classifier reading captions and hashtags — it's a visual model that watches the video (or looks at the image) directly: what product is physically shown on screen, how it's presented, what's said about it, and when. That visual read is then checked against several reference points at once:

• Product Center — the product's own image and description, so the agent can judge whether what's shown actually matches the real product, not just whether a product was mentioned by name
• The creator brief — from the collaboration this content is tied to (see Understanding Campaigns), so scoring reflects whether the content actually delivered what was briefed
• Brand Guidelines — your company's content and brand standards, set in Settings, which is the primary reference for Brand Safety specifically
• Patterns in your own tracked content — what "good" already looks like across your library, for Content Quality

Where scores show up

Scores appear directly on each post in Content Library — both the Content/Detail view (see Reading a Content/Detail page for the full breakdown, including where the visual model's read shows up) and as columns in Content/Overview (see Content Library: folders, review, and AI scores). They're also what other agents and features build on: the Content Tagging Agent uses Product Fit signal as part of its matching, and a low Brand Safety score is worth checking before content goes into a paid campaign or a Website Widget.

Reading a scoring pattern, not just one number

A single low score on one post usually isn't meaningful — creators vary, and not every post needs to be a home run. What's worth acting on is a pattern: a creator whose Hook scores are consistently low across many posts, or a product category that consistently scores low on Product Fit regardless of which creator posts about it.