> ## Documentation Index
> Fetch the complete documentation index at: https://docs.usequotable.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Read the Agent Analytics dashboard

> What each figure, tab, filter, alert and recommendation on the Agent Analytics page measures.

Open a brand and select **Agent Analytics**. The page is built from the accepted agent events of every source connected to that brand.

## The opening line and the four figures

The page opens with one sentence that says what happened in the selected range, then four figures. Each compares the range with the previous range of the same length.

| Figure | What it counts |
| - | - |
| **Answer visits** | Times an AI assistant opened a page to answer someone's question. |
| **Index crawls** | Visits from crawlers that build the search index AI answers draw on. |
| **Training crawls** | Visits from crawlers that collect pages to train models. |
| **All AI visits** | Every request from a recognized AI system. |

Select a figure to chart it. The chart draws the range as a solid line against the previous range as a dashed one. **Compare AI systems** splits the line by AI company, and **Day / Week** changes the grouping on ranges of 30 days or more.

Under the chart, a row gives the share of visits that were verified, the share that failed, the slowest 5% response time, and the number of pages and bots seen. Only requests from recognized AI systems are counted. Human visits that arrive from AI answers are a separate dataset on [AI Referrals](https://docs.usequotable.com/agent-analytics/ai-referrals.md).

## Filters

The range and the **Filters** menu narrow every tab at once. A filter you set appears beside the menu and can be removed there.

* **Range**: last 7, 14, 30 or 90 days.
* **AI system**, **Bot**, **Visit type** and **Verification**.
* **Response**: 2xx, 3xx, 4xx or 5xx.
* **Path**: pages whose path starts with what you type, for example `/blog`.
* **Country**: a two-letter country code, where the source provides country.
* **Source**: one connected source, when the brand has more than one.
* **Timezone**: the reporting timezone for daily buckets. UTC unless you change it.

## Tabs

> [tab] Overview
>
> * The opening sentence, the four figures and the trend chart.
> * **Needs attention**: active alerts (see below).
> * **AI systems**: visits per AI company, as a share or as totals.
> * **Visit types**: answer visits, index crawls and training crawls side by side.
> * **Top pages**: the pages AI assistants open most. Select a page to narrow the whole view to it.
> * **Failed requests**: the pages that most often return an error to AI systems.

> [tab] AI systems
>
> One row per AI company, with all visits and the change against the previous range, answer visits, index crawls, training crawls, the verified share, errors and when it was last seen. Open a row to see each of its bots. Switch to **Bots** for a flat list.

> [tab] Pages
>
> One row per page: answer visits, index crawls, training crawls, all visits, the number of AI systems, errors, the slowest 5% response time and when it was last seen. Select a column title to sort the whole table by it, search by path, or download the rows as CSV. Links open the same page in Page Tracker and, when AI Referrals is enabled, in [AI Referrals](https://docs.usequotable.com/agent-analytics/ai-referrals.md).
>
> A citations column can appear beside the visit columns. It comes from your citation tracking, is an all-time count, and shares no denominator with the visit columns. A crawl doesn't imply a citation.

> [tab] Crawlability
>
> * **robots.txt audit**: checks which AI agents your current robots.txt allows at the site root, lists declared sitemaps and shows changes since the previous audit. It runs on demand, from our network, against your brand's own domain.
> * **Test a page**: fetches one page and checks for problems agents run into, such as content that needs JavaScript or a challenge page, missing structured data, `noindex` or canonical issues, and thin internal linking.
> * What to fix is not listed here. Errors AI systems hit on real visits become items on the [Opportunities](https://docs.usequotable.com/actions/opportunities.md) board, next to the [site checks](https://docs.usequotable.com/actions/site-checks.md).
> * **Failed requests**: URLs that returned 4xx or 5xx to recognized AI systems.
>
> Diagnostics are limited to 30 runs per hour per brand, because each run fetches your site several times.

> [tab] Logs
>
> The newest accepted requests with time, AI system and bot, visit type, path, HTTP status, response time, verification and source. Search by path, page back with **Older**, or download the rows as CSV. Turn on **Live** to refresh the list every 10 seconds. Dropped records are not listed here.

> [tab] Sources
>
> Each connected source with its status, last delivery, last AI visit and visits in the range, with **Rotate token** and **Disconnect**. **Connect source** adds another.

## Alerts

Alerts are deterministic: each one states its threshold, its baseline and what resolves it. Active alerts show in **Needs attention** and are sent as notifications to the workspace owners and to members with access to the brand. An alert that stays active is sent again at most once every 24 hours.

| Alert | Fires when |
| - | - |
| Source error, disconnected or delayed | A source's status is error, disconnected or delayed. |
| Crawler failure rate is elevated | 10% or more of agent requests fail (critical at 25%). |
| Crawler response latency is elevated | p95 response time is 2,000 ms or more. |
| Crawler activity dropped or spiked | Requests change by 50% or more against the previous equal window, with at least 50 baseline requests. |
| New agent | An agent appears that wasn't seen in the previous window. |
| robots.txt policy changed, Sitemap changed | Two consecutive audits differ at warning level. |
| Quota | A source reaches 80% of its daily quota (critical at 100%). |

## Recommendations

Recommendations come from what agents actually experienced and from the synthetic checks. Examples:

* **Unblock**, **Repair** or **Stabilize** a path that returns errors to agents.
* **Reduce latency** on a slow page.
* **Agents fail across** a folder where at least a quarter of requests fail.
* **Publish an explicit robots.txt policy**, **Review AI search crawler blocks**, **Declare a sitemap in robots.txt**, **Fix the declared sitemap**.
* **Crawled, not cited**, or **not fetched by any agent**, by comparing crawls with your citation data.
* **Serve the page without JavaScript or a challenge**, **Add structured data**, **Link to related pages**, from the page test.

None of these promises a citation or a referral. They remove reasons an agent can't read your page.

## How fresh is the data

* Most push connectors deliver within a minute. Akamai DataStream 2 delivers every 30 seconds to a few minutes.
* Timestamps come from your logs, not from arrival time. A record older than seven days, or more than five minutes in the future, is dropped as `timestamp_out_of_range`.
* The Logs tab and the figures only include accepted events.
* A site no AI system has visited for a while is quiet, not delayed. The page says data is delayed only when accepted visits are taking longer than usual to appear.

## Next steps

> [card] [How verification works](https://docs.usequotable.com/agent-analytics/how-verification-works.md)
>
> The evidence behind verified, unverified and spoofed.

> [card] [Troubleshooting](https://docs.usequotable.com/agent-analytics/troubleshooting.md)
>
> When numbers look wrong.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.