29.9% of Downloads of Tools That Feed Web Pages to AI Came From Automated CI Pipelines in 2026, Up From 21.1% in 2024

Summary

Picture an HVAC contractor who updates his service area page on Monday. By Tuesday, several AI tools have already fetched it. Nobody asked them to; their software runs on a schedule and was rebuilt and tested overnight.

Tools that feed web pages to AI models, from web-to-text converters to browser agents, are installed from the Python Package Index. When a machine downloads one of them inside a continuous integration (CI) system, the log records a CI flag. That flag is a direct, if partial, measure of how much of this software runs unattended as part of an automated build.

In the Q3 2026 sample days, 29.9% of downloads of 23 web-to-AI packages carried the CI flag, up from 21.1% two years earlier. Other signals point the same way: most downloads now come through an installer favored by automated builds, and weekend activity is catching up with weekdays.

What we measured

We queried the public PyPI download log in Google BigQuery (bigquery-public-data.pypi.file_downloads), which records one row for every file downloaded from the Python Package Index. To keep the query affordable we sampled the first seven days of every month from January 2022 through September 2026, 399 days in total, and summed downloads by month, package, country, CI flag, installer and weekday versus weekend into one working table. A quarter in this study therefore means 21 sampled days (days 1 to 7 of each of its three months), not a full quarter. We removed downloads made by mirroring and proxy tools (bandersnatch, Nexus, Artifactory, devpi), which copy packages in bulk rather than install them. To control for the growth of PyPI itself we also pulled total daily download counts for the whole index from the table's partition metadata for the same sampled days.

The group combines the 14 web-to-text packages, 8 search API packages and the browser-use agent package from our related studies. A download is CI-flagged when the installing machine reports a known CI environment variable. CI data starts in March 2024, so comparisons begin in Q3 2024. Production servers and containers often do not set CI variables, so the CI share is a lower bound for automated downloads.

The CI share has risen since 2024

CI-flagged share of downloads by monthly sample week (days 1 to 7), since the field became available in March 2024.

The share moved between 14.6% and 48.6% across these weeks. It was 48.6% in the September 2026 sample week, the most recent in our data.

Search APIs are the most pipeline-heavy

CI share and downloads by tool type, Q3 2026 sample days.

The browser agent category had a low CI share (5.9%) but 76.9% of its downloads came through uv, which suggests installs in sandboxes and containers that do not identify as CI.

Installers and timing point to unattended use

Installer mix for the whole group at three points in time.

uv is a fast installer widely used in containers, CI and AI agent tooling, though developers also use it by hand. Weekend downloads rose from 50.5% to 70.0% of the weekday pace, close to the 65.5% for PyPI overall in Q3 2026.

What this means

A growing share of the software that reads your site for AI is built, tested and redeployed automatically. Once it is running, it checks pages on its own schedule, so changes to your site reach these tools quickly, and so do errors. A broken page, an accidental noindex or a firewall rule that blocks good crawlers can spread into AI answers before anyone notices.

Treat site changes as if machines will read them the same day, and monitor what AI assistants say about you. A free AI visibility audit is a good place to start, and our LLM optimization plan then rechecks those answers each month after your site changes.

Limits

A download is not a user. One developer, one Docker build or one CI job can download the same package many times, and a single busy pipeline can move a package's numbers sharply from month to month.

Figures cover the first seven days of each month only (21 days per quarter), so they are sample totals, not full-month or full-quarter totals. Growth rates compare like-for-like sample periods.

PyPI as a whole grew from 12.1 billion downloads in the Q3 2022 sample days to 115.0 billion in the Q3 2026 sample days (9.5x), so part of every package's growth is the index growing. We report share of all PyPI downloads where it matters.

PyPI-wide downloads in the September 2026 sample week were 38.1% lower than in the August 2026 sample week (28.7 billion vs 46.4 billion), so September figures for individual packages should be read with that drop in mind.

The CI flag is only populated in this dataset from March 2024, and it only catches environments that announce themselves as CI. Automated downloads that do not set a CI variable are counted as non-CI.

Country codes come from the downloading IP address. Cloud servers sit mostly in the United States, so country splits describe where machines run, not where developers live.

CI flags undercount automation; servers, containers and agent sandboxes that do not set CI variables count as non-CI.

Installer choice is a weak signal on its own, because developers also use uv interactively.

Frequently asked questions

Related studies

• Downloads of the tools that read websites for AIMCP Downloads Now Exceed 10 Agent Frameworks Combined
• Downloads of the tools that read websites for AIAI Agent Framework Downloads Topped 108M in One Week
• Downloads of the tools that read websites for AIWeb-to-Text Tool Downloads Grew 71x Since 2022
• Downloads of the tools that read websites for AIPDF-to-Text Downloads for AI Rose 65x in Four Years