AI Visibility Analytics

AI Visibility gives you two related but separate signals:

What AI crawlers fetch from your site
Which of those fetched pages later attract AI-referred human visits

That distinction matters. A crawler fetch is not the same thing as a human referral, and HitKeep keeps those signals separate instead of collapsing them into one number.

HitKeep AI visibility overview with fetch KPI cards, filters, and fetch volume over time — Overview of the AI Visibility page: fetch KPIs, assistant/operator filters, and crawl demand over time.

What the dashboard shows

The AI Visibility page is split into two layers:

Fetch analytics with KPI cards for total fetches, unique paths, unique assistants, 4xx and 5xx rates, response time, and bytes served
Correlation analytics with summary KPIs plus full-width tabbed views for citation yield, opportunity pages, and failure hotspots

The page supports filtering by assistant, operator family, and resource type so you can answer questions like:

What is OpenAI fetching most often?
Which document fetches later attract AI-referred visits?
Where are Anthropic or Perplexity fetches hitting 404s?

HitKeep AI visibility correlation card with summary KPIs and tabbed citation yield, opportunity pages, and failure hotspots tables — Correlation analytics use full-width tabs so each data-dense table stays readable instead of collapsing into narrow columns.

What the correlation report shows

The AI fetch correlation report is a directional report.

The fetch side comes from server-side AI fetch records collected through POST /api/sites/{id}/ingest/ai-fetch.
The visit side comes from normal human pageviews where the referrer matches a known AI assistant such as ChatGPT or Perplexity.
Correlation happens when a fetched path later receives an AI-referred visit within the selected window.

This is intentionally not strict attribution. It is meant to answer:

“When these AI bots fetch a page, does that page later generate AI-origin traffic?”

It does not claim:

“This exact GPTBot fetch caused this exact ChatGPT referral.”

Filter semantics

Assistant filters such as assistant_name and assistant_family apply to the fetch side only.

That means:

filtering to OpenAI limits the fetched rows to OpenAI bots
correlated visit counts still include any later AI-referred human visits on the same path

This is the intended product behavior. It keeps the report focused on downstream visibility rather than implying deterministic bot-to-referrer attribution.

Metrics in the report

GET /api/sites/{id}/ai-fetch/correlation returns four sections:

summary: total fetches, fetched paths, correlated paths, later AI-referred visits, and uncorrelated fetches
citation_yield: paths with the strongest fetch-to-visit payoff
opportunity_pages: heavily fetched pages with weak downstream AI traffic and/or elevated errors
failure_hotspots: assistant and path-prefix combinations with concentrated 4xx/5xx issues

Example request

curl "https://your-hitkeep.example/api/sites/{id}/ai-fetch/correlation?\
from=2026-03-01T00:00:00Z&to=2026-03-31T23:59:59Z&\
assistant_family=OpenAI&window_days=30" \
  -H "Authorization: Bearer YOUR_API_TOKEN"

Example response

{
  "summary": {
    "total_fetches": 124,
    "fetched_paths": 38,
    "correlated_paths": 11,
    "ai_referred_visits": 27,
    "uncorrelated_fetches": 73
  },
  "citation_yield": [
    {
      "path": "/guides/ai-visibility",
      "assistant_name": "GPTBot",
      "fetch_count": 8,
      "ai_referred_visits": 5,
      "citation_yield_pct": 62.5
    }
  ],
  "opportunity_pages": [
    {
      "path": "/pricing",
      "fetch_count": 12,
      "ai_referred_visits": 1,
      "error_requests": 3,
      "error_rate_pct": 25
    }
  ],
  "failure_hotspots": [
    {
      "assistant_name": "ClaudeBot",
      "path_prefix": "/docs",
      "total_requests": 9,
      "error_requests": 4,
      "error_rate_pct": 44.44
    }
  ]
}

Recommended interpretation

Use the report like this:

High fetches + high AI-referred visits: pages AI systems appear to find and surface successfully
High fetches + low visits: pages worth improving for titles, structure, freshness, and reliability
High error hotspots: technical issues that may reduce AI visibility even when crawlers are interested

The safest mental model is:

fetches show AI discovery, referrals show downstream human demand, and correlation shows where those two signals overlap.

Demo data

If you seed a local demo instance, the AI Visibility page is designed to open with realistic assistant mixes, resource-type filters, and populated correlation tabs right away instead of an empty shell.