Guide · Docker and Actions

How do AI tools recommend Docker images and GitHub Actions? Docker Hub and Marketplace pages, tested

Updated 2026-11-12 · kagentx

Short answer

Tested on 2026-11-12: Docker Hub image pages and GitHub Marketplace Action pages are both server-rendered, with the full README in the raw HTML; the meta description is Docker Hub's short description and the action.yml description respectively, both sentences you write yourself. Docker Hub's robots.txt has no disallow rules, and GitHub's rule for AI crawlers is Crawl-delay: 1 with no site-wide block. Docker specifies a short description of at most 100 characters and at most 3 categories, and GitHub requires Action names to be unique. Whether AI cites these two kinds of page, we have no evidence, so it is unverified.

Are Docker Hub image pages and Marketplace Action pages server-rendered?

Yes. On 2026-11-12 we fetched 3 pages with curl (no script execution, User-Agent Mozilla/5.0, no AI crawler simulated), and the body text is in the raw HTML in all of them:

PageHTTP status and sizetitle and meta descriptionVisible text (scripts removed)
hub.docker.com/_/nginx200, 524,903 bytestitle “nginx - Official Image | Docker Hub”; description is the short description “Official build of Nginx.”; has a canonicalAbout 13,737 characters: full README, docker pull and docker run
hub.docker.com/r/grafana/grafana200, 338,679 bytestitle “grafana/grafana - Docker Image”; description is the short description “The official Grafana docker container”; has a canonicalAbout 1,614 characters: full README (full_description in the API is only 569 characters), tag summary, docker pull, pulls this week
github.com/marketplace/actions/checkout200, 306,881 bytestitle “Checkout · Actions · GitHub Marketplace · GitHub”; description is identical to the description in action.yml; has a canonicalAbout 14,495 characters: README Usage, input descriptions, examples

None of the three pages has JSON-LD. Crawlers that don't run scripts can still read these pages; for whether crawlers run JavaScript in general, see Do AI crawlers render JavaScript?.

One key fact: on both kinds of page the meta description is a sentence you write yourself (Docker Hub's short description, action.yml's description), not the opening of the README. You need to write that sentence well.

What do Docker Hub's robots.txt and API look like?

CheckResults (2026-11-12)
hub.docker.com/robots.txt200, 98 bytes: User-agent: * followed by an empty Disallow, plus one Sitemap line pointing to sitemap-index.xml on cloudfront; no AI crawler is named
hub.docker.com/sitemap.xml200, 1,841 bytes, an index whose child files are sitemap-001.xml.gz and so on on cloudfront, the first few with lastmod 2024-01-17; the content of the child files is unverified
hub.docker.com/v2/repositories/library/nginx/200, 17,842 bytes of JSON, no login needed; fields include description (short description), full_description (16,434 characters), star_count 21,421, pull_count 13,582,768,795, categories, last_updated
docs.docker.com/robots.txt200, 116 bytes: one extra line, Content-Signal: ai-train=yes, search=yes, ai-input=yes; this is a declaration for the docs site, not for Docker Hub

On the same endpoint, grafana/grafana has a pull_count of 5,386,765,446 and three categories: Security, Internet of things, Monitoring & observability. These numbers were taken on 2026-11-12 and will change.

What does Docker Hub officially specify for descriptions and categories?

The repository information page on docs.docker.com (original text read on 2026-11-12; the page has no date):

ItemOfficial statement
Short descriptionUp to 100 characters; shown in docker search results
OverviewWith automated builds enabled, it can be synced from the repo's README.md; basic Markdown is supported; the page gives no length limit
CategoriesUp to 3, representing the image's main use cases; new repositories are not assigned any automatically; automatic assignment applies only to Verified Publisher and Docker-Sponsored Open Source projects
Being foundThe page says clear repository information helps others find and use your image, and categories make it easier for users to browse by problem domain

The vendor doesn't say how these fields affect search ranking or AI citation, so we have no evidence on that and it is unverified. For how to write the README, see Will AI search cite my GitHub README?.

Where does the data on a GitHub Marketplace Action page come from?

From action.yml in the repo root. In actions/checkout's action.yml (5,144 bytes), name is Checkout and description is Checkout a Git repository at a particular version, and each input below has its own description and default. The Marketplace page's meta description is that description sentence.

name: 'Csv Lint'
description: 'Check CSV files for broken rows and fail the workflow'
inputs:
  path:
    description: 'Glob of CSV files to check, for example data/**/*.csv'
    default: '**/*.csv'
runs:
  using: 'node20'
  main: 'dist/index.js'
branding:
  icon: 'check-circle'
  color: 'green'

The above is a writing example: let the description say in one sentence what it does and when it fails, and give each input an example value, because this text appears on the page and in the README.

GitHub's publishing docs (original text read on 2026-11-12; the page has no date):

ItemOfficial statement
NameThe name in action.yml must be unique: it can't match an existing action on the Marketplace, nor a user or organization name (unless published by that owner)
CategoriesPick one primary category (it should be the one that helps people find you) and optionally one secondary category
Repository layoutOnly one action.yml or action.yaml per repository root; ones in subfolders are not listed automatically
PublishingPublish a release; two-factor authentication is required; it is listed immediately once the conditions are met, and GitHub does not review it
Not specifiedRequirements for description and branding, and search ranking, are not specified

In GitHub's robots.txt, GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot and Perplexity-User are in one group: Crawl-delay: 1, no Disallow: /; under /marketplace/ only URLs with ?tab=readme and ?tab=transparency have rules, and Action pages themselves are not matched (this is the result of reading the rules, not of real crawler visits).

How do you make images and Actions easier to find, step by step?

  1. 1. Write the short description and description as meta descriptions

    Docker Hub caps it at 100 characters, while action.yml's description has no limit but becomes the meta description: say what it does and who it's for, without marketing adjectives.

  2. 2. Put copyable commands at the top of the README

    For an image, put docker pull and one docker run example; for an Action, put the uses: line and a complete workflow snippet. Both pages server-render the README for crawlers.

  3. 3. Pick categories on Docker Hub and a primary category on the Marketplace

    Docker Hub allows at most 3 and does not assign any automatically to new repositories, so pick them yourself; on the Marketplace, pick one primary category that helps people find you.

  4. 4. Keep the name unique and consistent with the official site

    The Action's name must be unique per the docs; keep the image name consistent with the official site and README, so AI doesn't see different names on different pages.

  5. 5. Write one page on the official site and link to both pages

    On the official site, write out the purpose, comparisons, versions and changelog, then link to the Docker Hub and Marketplace pages. For how to write it, see How should developer docs be written so AI can understand and cite them? and How do open source projects get recommended by AI search? Six places beyond the README.

  6. 6. Test again after a few weeks with the same questions

    Ask ChatGPT and Perplexity “which Docker image is best for X” and “is there a GitHub Action for X”, asking each question 3 times, and note whether you are mentioned and which URLs are cited. For images of MCP servers see How do you get an MCP server recommended by AI and listed in connector directories?, for package registry pages see How do packages on npm and PyPI get recommended by AI?, and for how to write robots.txt see How do you write robots.txt for AI crawlers?.

If you publish a Home Assistant custom integration that is also hosted on GitHub, the HACS requirements and the public pages we checked are in How do Home Assistant custom integrations get found by AI search?.

FAQ

Can crawlers that don't run scripts read a Docker Hub image page?

Yes. On 2026-11-12 we fetched the nginx and grafana/grafana pages with curl; both returned 200, and with scripts stripped they have about 13,737 and 1,614 visible characters, including the full README, the docker pull command and pull counts.

What is the meta description of a Docker Hub page?

It is the repository's short description. The nginx page has “Official build of Nginx.” and the grafana/grafana page has “The official Grafana docker container”. The docs say the short description is at most 100 characters, so it has to say everything in one sentence.

Does Docker Hub's robots.txt block AI crawlers?

No. Tested on 2026-11-12: hub.docker.com/robots.txt is 98 bytes: an empty Disallow under User-agent: *, plus one Sitemap line. The content of the child sitemap files is unverified.

Where does the meta description of a GitHub Marketplace Action page come from?

From the description in action.yml. The meta description of the actions/checkout page is identical to the description in action.yml: Checkout a Git repository at a particular version.

Can the README be read on an Action page?

Yes. On 2026-11-12 we fetched marketplace/actions/checkout: the page is 306,881 bytes and has about 14,495 visible characters with scripts stripped, including the Usage section and input descriptions.

What does GitHub's robots.txt say about AI crawlers?

8 AI-related User-Agents (GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User) are in one group with Crawl-delay: 1 and no Disallow: /. Under /marketplace/ only URLs with ?tab=readme and ?tab=transparency have rules, and Action pages themselves are not matched.

What are the naming requirements on the Marketplace?

The official docs say the name in action.yml must be unique: it can't match an existing action on the Marketplace, nor a user or organization name (unless published by that owner).

How does Marketplace search rank results?

The official docs don't describe search ranking or keyword rules; they only say the primary category helps people find your action. Whether AI cites Marketplace pages, we have no evidence, so it is unverified.

Sources

kagentx

Can crawlers read your pages?
Let the evaluation tell you

The evaluation checks whether crawlers can get in and whether the text is served directly, then asks ChatGPT, Perplexity and Claude (with web search on) your buyers' questions and sees whether you are mentioned. The Basic plan is $1, and results usually take 2–3 minutes.