This study examines web-server requests that identified themselves with documented AI-related user agents during 14 complete UTC days, from July 5 through July 18, 2026. The headline total includes only approved agents classified as user-triggered, search/indexing, or training activity.
A request is not a person, unique visit, citation, lead, order, or revenue event. Classification is based on self-identified user agents and published vendor documentation. Teal Packaging did not perform IP verification, reverse-DNS validation, or bot-authentication checks, so user-agent spoofing remains possible.
Time series
Daily eligible AI crawler requests
Requests varied sharply across the window, including two high-volume days on July 14 and July 15. The chart separates the three activity categories that make up the headline total.
Daily eligible AI crawler requests
Requests from approved self-identified agents, July 5 to July 18, 2026 UTC. Categories sum to the headline total.
| UTC date | User-triggered | Search/indexing | Training | Eligible total |
|---|---|---|---|---|
| 2026-07-05 | 330 | 174 | 457 | 961 |
| 2026-07-06 | 857 | 179 | 93 | 1,129 |
| 2026-07-07 | 547 | 336 | 73 | 956 |
| 2026-07-08 | 726 | 355 | 28 | 1,109 |
| 2026-07-09 | 671 | 720 | 59 | 1,450 |
| 2026-07-10 | 594 | 399 | 179 | 1,172 |
| 2026-07-11 | 536 | 190 | 25 | 751 |
| 2026-07-12 | 886 | 49 | 29 | 964 |
| 2026-07-13 | 530 | 62 | 352 | 944 |
| 2026-07-14 | 3,407 | 517 | 2,304 | 6,228 |
| 2026-07-15 | 5,826 | 550 | 556 | 6,932 |
| 2026-07-16 | 861 | 509 | 28 | 1,398 |
| 2026-07-17 | 578 | 118 | 39 | 735 |
| 2026-07-18 | 596 | 948 | 95 | 1,639 |
Headline population
What makes up the headline total?
User-triggered requests were the largest eligible category. Categories describe the documented role of an agent label, not the proven purpose of an individual request.
Eligible request share by documented activity category
Share of 26,368 eligible requests across the 14-day UTC window.
- User-triggered16,945, 64.26%
- Search/indexing5,106, 19.36%
- Training4,317, 16.37%
| Category | Requests | Share | Included agent labels |
|---|---|---|---|
| User-triggered | 16,945 | 64.26% | ChatGPT-User, Claude-User, Perplexity-User, Amzn-User |
| Search/indexing | 5,106 | 19.36% | OAI-SearchBot, Claude-SearchBot, PerplexityBot, Amzn-SearchBot |
| Training | 4,317 | 16.37% | GPTBot, ClaudeBot |
Agent registry
Eligible requests by approved user-agent label
Each label is matched with literal, case-insensitive boundaries and assigned only when exactly one distinct approved token appears in the logged user-agent field.
Eligible requests by approved user-agent label
Each bar is a self-identified label included in the headline population. Zero-count approved labels remain in the registry and table.
| Approved agent label | Vendor | Activity category | Requests | Share | Official documentation |
|---|---|---|---|---|---|
| ChatGPT-User | OpenAI | User-triggered | 16,846 | 63.89% | Read OpenAI documentation |
| OAI-SearchBot | OpenAI | Search/indexing | 4,726 | 17.92% | Read OpenAI documentation |
| ClaudeBot | Anthropic | Training | 3,966 | 15.04% | Read Anthropic documentation |
| PerplexityBot | Perplexity | Search/indexing | 380 | 1.44% | Read Perplexity documentation |
| GPTBot | OpenAI | Training | 351 | 1.33% | Read OpenAI documentation |
| Claude-User | Anthropic | User-triggered | 98 | 0.37% | Read Anthropic documentation |
| Perplexity-User | Perplexity | User-triggered | 1 | 0.00% | Read Perplexity documentation |
| Amzn-SearchBot | Amazon | Search/indexing | 0 | 0.00% | Read Amazon documentation |
| Amzn-User | Amazon | User-triggered | 0 | 0.00% | Read Amazon documentation |
| Claude-SearchBot | Anthropic | Search/indexing | 0 | 0.00% | Read Anthropic documentation |
Requested content
Where eligible AI agents requested pages
Every eligible headline request is assigned to one mutually exclusive page type. The classification describes URL structure only and does not measure page quality, citation, indexing, or business performance.
Headline page-type distribution
Page-type distribution within 26,368 eligible requests.
| Page type | Requests | Share | Classification rule |
|---|---|---|---|
| Other content | 9,847 | 37.34% | Eligible content path not assigned by an earlier page-type rule. |
| Product | 7,787 | 29.53% | First match: normalized path begins /product or /product/. |
| Location | 3,904 | 14.81% | First match: approved custom-boxes, custom-packaging, or packaging-supplies location slug. |
| Homepage | 2,225 | 8.44% | Normalized path is exactly /. |
| Comparison/reference | 1,780 | 6.75% | Final slug contains -vs-, matches best-*-alternatives, or contains calculator, chart, guide, statistics, or selector as a hyphen-delimited term. |
| Blog | 717 | 2.72% | First match: normalized path begins /blog or /blog/. |
| Shop | 108 | 0.41% | First match: normalized path begins /shop or /shop/. |
Secondary context
Multipurpose crawlers observed
Secondary context: Applebot and Amazonbot
These 13,932 requests are shown separately and do not contribute to the 26,368 headline total.
| User-agent label | Requests | Included in headline? | Reason for secondary treatment | Official documentation |
|---|---|---|---|---|
| Amazonbot | 12,772 | No | Multipurpose token. The purpose of an individual request cannot be inferred from the user-agent label alone. | Read Amazon documentation |
| Applebot | 1,160 | No | Multipurpose token. The purpose of an individual request cannot be inferred from the user-agent label alone. | Read Apple documentation |
Reproducibility
Methodology and limitations
How this study classified requests
- The study window is 00:00:00 UTC on July 5, 2026 through 23:59:59 UTC on July 18, 2026, inclusive.
- Requests were grouped from web-server logs using self-identified user-agent strings. They were then matched to an approved registry with official vendor documentation.
- The headline total contains only eligible user-triggered, search/indexing, and training agents. 40,300 is a secondary context total that also adds multipurpose Applebot and Amazonbot requests.
- Eligible requests use GET or HEAD, return status 200, and target a content path after the published path-normalization and exclusion rules. HEAD requests are tallied separately in the downloadable JSON.
- A logged request can be automated or spoofed. The analysis does not prove the requester identity or purpose because it does not include IP verification, reverse-DNS checks, or bot-authentication validation.
- Counts describe requests, not people, unique visitors, citations, leads, orders, conversions, revenue, or causal effect.
- Page type reflects the requested route classification, not a judgment about a page's quality, indexing, citation, or commercial value.
How to cite this research
Teal Packaging. (2026, July 19). AI Crawler Statistics for Teal Packaging, July 2026. https://tealpackaging.com/stats/ai-crawler-statistics-2026/. Study window: July 5 to July 18, 2026 UTC. Accessed July 19, 2026.
Please cite the page and the matching dataset version together when using a chart, table, or request count.
Open data
Download the sanitized research data
Dataset version 2026-07-19.v1. Downloads contain aggregated or sanitized request data only. They do not include IP addresses, cookies, query strings, or other direct identifiers.
- Download sanitized JSON SHA-256 6ee1bc24cb09f3873211b71da675b4b19f6d97af973007c81848fb64d35cd938
- Download sanitized CSV SHA-256 95faf4c892c638d53aca54a74e112cac70ac62de863f15fa6ad0f8144344fb9c
- Read the field dictionary and classification registry SHA-256 e22f9d74f5ee1145ec0a54842a94fbbe9ff4f384c10c85c22ae67a93aaa3a451
Questions
Frequently asked questions
What does the headline total measure?
It measures 26,368 logged requests from approved self-identified user agents during 14 complete UTC days. It is not a people, visitor, citation, lead, order, or revenue count.
Why are Applebot and Amazonbot not in the headline total?
They are multipurpose crawlers. Their user-agent strings alone do not establish an AI-related purpose for a given request, so they are shown only as secondary context.
Can a user agent be spoofed?
Yes. This study classifies requests by self-identified user agents and official vendor documentation. It does not perform IP verification, reverse-DNS validation, or bot-authentication checks.
What pages do the requests represent?
The page-type analysis groups requested routes using the published classification rules. It does not show whether a page was indexed, cited, read by a person, or commercially successful.
Can I reuse the data or charts?
Use the downloadable sanitized JSON or CSV and cite the page plus dataset version. Keep the study window, request definition, and limitations attached to any reused number.
Documentation
Sources
Agent classifications use current official vendor documentation accessed July 19, 2026. The first-party counts come from the versioned sanitized dataset and independently validated daily aggregates.
- OpenAI crawler documentation: ChatGPT-User, OAI-SearchBot, and GPTBot.
- Anthropic crawler documentation: Claude-User, Claude-SearchBot, and ClaudeBot.
- Perplexity crawler documentation: Perplexity-User and PerplexityBot.
- Amazon crawler documentation: Amzn-User, Amzn-SearchBot, and Amazonbot.
- Applebot documentation: Applebot search and data-use controls.
- Google crawler documentation: Google-Extended exclusion rationale.
- Teal Packaging sanitized request dataset, version 2026-07-19.v1.
- Teal Packaging field dictionary and agent registry.