> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.middesk.com/online-presence/risky-keywords/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.middesk.com/_mcp/server.

# Detect risky keywords

> Use Middesk risky keyword detection to scan a business's web presence for terms tied to prohibited or high-risk activity

Risky keyword detection scans a business's web presence for terms tied to prohibited or high-risk activity. [Industry classification](/online-presence/industry-classification) verifies what a business does, but a business in an acceptable category can still sell prohibited products, and that activity often shows up in its own marketing or in customer reviews before anywhere else. Payment processors and platforms use risky keyword results to catch this conduct early, including the brand-damaging and illegal activity that card network enforcement programs like Mastercard BRAM (Business Risk Assessment and Mitigation) penalize.

## How Middesk detects risky keywords

Risky keyword scanning runs as part of [web presence analysis](/online-presence/web-presence). Middesk scans the business's website along with third-party profile content, including posts and reviews on platforms such as Google, Facebook, LinkedIn, Instagram, and Yelp.

Middesk groups matches into results by category. Each result includes:

* The `category` of risky keyword and an `explanation` of why the category appears in the results.
* A `severity` of `prohibited`, `needs_review`, or `ignore`, set by the rules configured for your account. Middesk provides a default rule set, and you can tailor the categories, keywords, and severities to match your policy.
* One or more `sources[]`, each with the `url` where the match was found, the `matched_keywords`, and a `text_snippet` of the surrounding content. The `source_type` field identifies where the content came from, and post and review sources also include a `posted_at` timestamp. See the [Website reference](/reference/website) for the full list of source types.

The following example shows one entry in `risky_keywords_results[]`:

**`Example risky keywords result`**

```json title="Example risky keywords result"
{
  "category": "cannabis",
  "explanation": "Multiple sources advertise THC vape products for sale.",
  "severity": "prohibited",
  "sources": [
    {
      "url": "https://www.instagram.com/p/aBc123XyZ45/",
      "matched_keywords": ["thc", "vape"],
      "text_snippet": "New drop: premium THC vape cartridges, available in store and online.",
      "source_type": "instagram_recent_post",
      "posted_at": "2026-06-14T18:22:05.000Z"
    },
    {
      "url": "https://www.example.com/shop",
      "matched_keywords": ["thc"],
      "text_snippet": "Shop our full range of THC products, glassware, and accessories.",
      "source_type": "website",
      "posted_at": null
    }
  ]
}
```

## How to use risky keyword results

#### Read the results

Risky keyword results appear under `risky_keywords_results[]` on the business's [website object](/reference/website). On Risk orders, results also appear at the top level of the business. A summary outcome appears in the `risky_keywords` [review task](/online-presence/web-presence/review-tasks#risky_keywords-review-task).

#### Act on severity

Severity maps each result to an action:

* `prohibited` findings indicate activity your program does not allow. Decline the business or escalate to your compliance team.
* `needs_review` findings warrant a closer look. Route them to manual review along with the evidence in `sources[]`.
* `ignore` findings are informational and need no action.

> **Get a demo**
>
> Contact your account manager or [contact sales](https://www.middesk.com/contact-sales) to inquire about access.

> Use Middesk risky keyword detection to scan a business's web presence for terms tied to prohibited or high-risk activity