<!-- The markdown copy of https://oppalerts.com/docs/mcp/llm_answers_content_phrases . All pages: https://oppalerts.com/docs/llms.txt -->

[Documentation](https://oppalerts.com/docs/) / [MCP](https://oppalerts.com/docs/mcp/) / llm_answers_content_phrases

# llm_answers_content_phrases

`POST https://mcp.oppalerts.com/mcp tool: llm_answers_content_phrases`

The wording the AI models use when they write about this market: the content phrases, scored across every model and run.

## What it does

The tool lists the wording that the AI models use when they write about this market: the content phrases. It scores each phrase across every model and every run. These phrases are writing vocabulary. They are never searched.

It needs a campaign.

## Connect

```bash
claude mcp add --transport http oppalerts https://mcp.oppalerts.com/mcp
```

The first call opens a browser window for the OppAlerts login. [Connect an AI Agent](https://oppalerts.com/docs/mcp/) has the details.

## Parameters

**`campaign`int, required, 1 to 4294967295**

The campaign id, from [campaigns_list](https://oppalerts.com/docs/mcp/campaigns_list). It must be a campaign of your account.

**`page`int, required, 1 to 1000**

The page to return, a whole number from 1 to 1000. A page holds 1000 items.

## Strict Rules

- A field that is not in the list above is refused. The result then has `isError` true, and its text starts with `bad_param`. The error names the field: "<field> is not a field of this tool." The old name of a renamed field is refused in the same way.
- A bad value is refused, and the error names the field. A bad value is a value of the wrong type, a number with a fraction for a whole-number field, a value out of its range, or a value that is not in its list. A required field that you do not send is refused too. The tool never changes a value that you send.
- An error about the format of a call, such as a refused field, names the field. Its message ends with one sentence: "Read the documentation at https://oppalerts.com/docs for the correct format." Every other error has only its own sentence.

## Example

### Call

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "llm_answers_content_phrases",
    "arguments": {
      "campaign": 3,
      "page": 1
    }
  }
}
```

### Response

Shown as it is in the source sample: one item.

```json
{
  "models_total": 6,
  "total": 412,
  "page": 1,
  "pages": 1,
  "items": [
    {
      "item": "book direct and skip the fees",
      "agreement_score": 2.11,
      "appearances": 6,
      "models": 5,
      "best_rank": 2,
      "forms": {
        "book direct and skip the fees": 6
      }
    }
  ]
}
```

### Response fields

- `models_total`: the number of AI models in the runs.
- `total`: the number of items in the whole list.
- `page` and `pages`: this page and the number of pages.
- `items`: the list, best score first. Each item has the fields below.

**`item`**

a content phrase.

**`agreement_score`**

The score of the item across every model and run.

**`appearances`**

How many times the models returned the item.

**`models`**

How many models returned the item. The response does not identify the models.

**`best_rank`**

The item’s highest rank in a single response.

**`forms`**

The ways the models wrote the item, each with its count. Items merge on normalized text, so several forms can share one item.

## Errors

An error returns as a tool result with `isError` true. Its text starts with the error code, except for `daily_limit`, whose text is only the sentence. The HTTP status shown for each code is its API status.

- `bad_param` (HTTP 400 in the API): a field is not a field of this tool, a required field is not sent, or a value has the wrong type or is out of its range. `campaign` and `page` are required, and `page` runs from 1 to 1000. The error names the field.
- `no_campaign` (HTTP 404 in the API): That campaign is not in this account. The same error comes for a campaign id that does not exist.
- `daily_limit` (HTTP 429 in the API): the account has used all its calls for this tool today. [account_calls_left](https://oppalerts.com/docs/mcp/account_calls_left) shows when the count resets.

This tool has its own daily count. A call through the API or MCP uses one call from this count. A web app page reading the same data does not count. It is not an index search, so it does not use the shared count of the `news_` and `prospects_` tools, and the error `one_at_a_time` does not apply to it.

## Same data in the API

```http
POST /v1/llm_answers/content_phrases HTTP/1.1
Host: api.oppalerts.com
Authorization: Bearer YOUR_TOKEN
Content-Type: application/json

{"campaign": 3, "page": 1}
```

See [POST /v1/llm_answers/content_phrases in the API](https://oppalerts.com/docs/api/llm_answers_content_phrases).

## Related pages

- [POST /v1/llm_answers/content_phrases in the API](https://oppalerts.com/docs/api/llm_answers_content_phrases)
- [Keyword Research in the web app](https://oppalerts.com/docs/app/content/keyword_research)
- [llm_answers_top_sites: The same shape for domains](https://oppalerts.com/docs/mcp/llm_answers_top_sites)
- [account_calls_left: the calls left today](https://oppalerts.com/docs/mcp/account_calls_left)
