pressfit.aiResearch
Pressfit.ai 30-Day Telemetry|November 2026

The AI Search Attribution Gap

Google Analytics saw about 6,500 human visits and almost no AI. Our firewall and edge logs saw AI and search crawlers read Pressfit.ai 17,630 times in the same 30 days.

GA4 Direct Share
99.38%

6,460 of about 6,500 human visits. One AI referrer showed up: Grok, 10 visits.

AI & Answer-Engine Bots
1.13x

7,321 requests, more than our human traffic.

AI + Search Crawlers
2.71x

17,630 requests. 3.49x (22,692) counting AdsBot and other Google services.

Live Human-Triggered Fetches
187

127 ChatGPT-User, 57 Claude-User, 3 ExaUser. GA4 showed zero from them.

The Browser Blind Spot vs. the Server Reality

Client-side analytics only count sessions where the tag fires and the referrer survives. Mobile AI apps, logged-in sessions and copied links drop the referrer, so AI-driven visits land in Direct. The edge sees every request.

What GA4 Showed (30 days)

About 6,500 human visits, almost all of them Direct.

Takeaway: a dashboard like this makes ChatGPT, Claude and Meta AI look like they send nothing.

What the Edge Logs Showed (same 30 days)

Human visits against AI bots, search crawlers and paid or secondary Google bots.

Ratios: AI bots 1.13x human visits, AI plus search crawlers 2.71x, all identified bots 3.49x (including 5,002 AdsBot and 60 other Google requests, kept out of the organic counts).

AI & Answer-Engine Bots: 7,321 Requests

Every one of these is an AI platform, retrieval API or agent reading Pressfit.ai. Almost none of them pass a referrer to Google Analytics.

Requests by AI Provider

Inside OpenAI's 1,789: 1,401 OAI-SearchBot (indexing), 132 GPTBot (training), 127 ChatGPT-User (live), 127 GPT Actions and 2 Operator (agents).

What Each Provider Is Doing

ProviderRequestsRole
Meta AI2,280 (31.1%)Indexing (1,764) and live web answers (516) for Meta AI
OpenAI1,789 (24.4%)Search indexing, live lookups and agent actions
Amazon1,234 (16.9%)Alexa, Rufus and Bedrock retrieval
Parallel613 (8.4%)AI web search behind agent APIs
Anthropic580 (7.9%)Model ingestion (523) and live Claude fetches (57)
Exa451 (6.2%)Neural search for AI apps and copilots
Perplexity346 (4.7%)Answer-engine indexing and citations
Common Crawl17 (0.2%)Open-web training corpus
DuckDuckGo AI11 (0.2%)Private AI answers

Search Crawlers That Feed AI Answers: 10,309 Requests

Classic search crawlers still ground Google AI Overviews and AI Mode, Microsoft Copilot and Apple Intelligence. Paid-ad verification bots are left out so ad sweeps don't inflate organic GEO counts.

Requests by Search Crawler

Googlebot alone made 6,486 requests (62.9%), followed by Bingbot at 1,393 (13.5%) and GoogleOther at 1,087 (10.5%).

Live Human Fetches vs. Background Crawling

1,401 OAI-SearchBot hits don't mean 1,401 people asked ChatGPT about you. Split the audited user agents into live, user-triggered fetches and background indexing or training.

Audited AI User Agents: 4,353 Requests

187 live fetches (4.30%), plus 129 autonomous agent actions, against 4,166 requests (95.70%) of search indexing (3,511) and background training (655).

Keep the Telemetry Honest

Verify the bot. User-agent strings can be spoofed. Check official IP lists, reverse DNS or Web Bot Auth signatures (IETF RFC 9421).

Check the status code. Count only HTTP 200 responses, so silent 403 WAF blocks and 499 timeouts don't look like healthy indexing.

Read the edge, not just the origin. CDN-cached hits never reach your origin server's logs.

Retrieval isn't revenue. A fetch proves a model read the page, not that it cited you or that a buyer converted. Tie it to analytics and CRM.

The Four-Layer AI Attribution Stack

Measuring SEO and GEO takes all four layers, using tools most marketing and engineering teams already own.

Layer 1: Site engineering

Server-rendered pages, a direct answer in the first 30% of the page, hub pages for fan-out subtopics and quarterly refreshes.

Layer 2: Edge telemetry

Cloudflare AI Crawl Control, AWS WAF Bot Control with CloudWatch, Vercel Observability or an Nginx log count.

Layer 3: Analytics

GA4's AI Assistant channel plus a custom regex channel, Search Console's AI Overviews and AI Mode filters, and direct-lift checks.

Layer 4: CRM

A required "How did you first hear about us?" field, tagged for AI mentions and joined back to GA4 conversions.

See where you stand in search and AI answers

Your Playbook maps where you and your competitors show up, then ranks every fix.

Get your Playbook →