Research & Reference

Find AI visibility reports for research and reference websites. Check bot access and look for source documents, dataset descriptions, versions and units.

How websites are scored

Signals

Each one you pick narrows the list: a site must have them all.

Agent capabilities

API catalog0Lists the site's public APIs in one standard place, so an agent can find and use them without reading the docs. A bonus: it earns a badge and never costs points.
auth.md0A simple file that tells AI agents how to sign in and get access. A bonus: it earns a badge and never costs points.
A2A agent card0Describes the site's own AI agent in the A2A format, so other agents can find it and hand it work. A bonus: it earns a badge and never costs points.
Agent Skills index0Lists reusable instructions the site publishes for AI agents, with descriptions and download locations. The index earns a badge; the skill downloads themselves are not verified.
Web Bot Auth0Publishes the keys the site's own bots sign their requests with, so other websites can verify them. The badge checks the keys; the bots' requests are not observed.
ACP commerce0Publishes how agents can connect to its Agentic Commerce Protocol integration. The badge confirms discovery metadata; purchases are not tested.

Standards

llms.txt1A short guide to the site written for AI models: what it is and which pages matter most.
robots.txt1The file that tells crawlers where they may go. Every other rule a site gives to bots starts here.
Sitemap1A list of the site's pages, so a bot sees everything that exists instead of guessing from links.
Descriptive llms.txt1The llms.txt describes what is behind each link, so a model knows which pages are worth reading.
llms-full.txt or .md pages1The content is also available as plain Markdown, in one llms-full.txt file or as separate .md pages. Models can read it without the HTML.
Markdown negotiation0Pages come back as clean Markdown when a bot asks for it: just the content, without menus, styling and scripts.
AI crawler rules1The site says which AI bots may visit, for training, search or answering questions. Blocking some is fine; what counts is making a choice.
Content Signals0The site says how its content may be used: for search, in AI answers or to train AI. Any mix of yes and no counts.
schema.org JSON-LD1Structured data on the homepage says what the site is, such as a company, a product or an article, so machines don't have to guess.
Semantic HTML1The page has a clear structure: a main content area, headings in order, a header and a footer. A bot can tell the content from the rest.
Crawlable navigation1Observed homepage navigation exposes usable page addresses and section targets in HTML. Destinations are not fetched by this check.
Named links and buttons1Links and buttons in the collected homepage HTML have names recoverable from text, labels or native alternatives, including icon controls.
Named form fields0Observed homepage fields have recoverable names and recognizable input semantics. Submission behavior is not tested.
Control roles and states0Homepage controls meet static checks for roles, required states and relationship targets, using native HTML semantics where available.
Typed links0The page links to versions made for machines, such as a Markdown copy or an API description, so an agent can go straight to them.
Canonical URL1Each page names the one address that should be indexed and cited, so an agent quoting it links to the right URL.
Working section links1Links to part of a page, such as a table of contents or a skip link, lead to that section instead of the top of the page.

Clear signals

Filters · 2 active

Signals

Each one you pick narrows the list: a site must have them all.

Agent capabilities

API catalog0Lists the site's public APIs in one standard place, so an agent can find and use them without reading the docs. A bonus: it earns a badge and never costs points.
auth.md0A simple file that tells AI agents how to sign in and get access. A bonus: it earns a badge and never costs points.
A2A agent card0Describes the site's own AI agent in the A2A format, so other agents can find it and hand it work. A bonus: it earns a badge and never costs points.
Agent Skills index0Lists reusable instructions the site publishes for AI agents, with descriptions and download locations. The index earns a badge; the skill downloads themselves are not verified.
Web Bot Auth0Publishes the keys the site's own bots sign their requests with, so other websites can verify them. The badge checks the keys; the bots' requests are not observed.
ACP commerce0Publishes how agents can connect to its Agentic Commerce Protocol integration. The badge confirms discovery metadata; purchases are not tested.

Standards

llms.txt1A short guide to the site written for AI models: what it is and which pages matter most.
robots.txt1The file that tells crawlers where they may go. Every other rule a site gives to bots starts here.
Sitemap1A list of the site's pages, so a bot sees everything that exists instead of guessing from links.
Descriptive llms.txt1The llms.txt describes what is behind each link, so a model knows which pages are worth reading.
llms-full.txt or .md pages1The content is also available as plain Markdown, in one llms-full.txt file or as separate .md pages. Models can read it without the HTML.
Markdown negotiation0Pages come back as clean Markdown when a bot asks for it: just the content, without menus, styling and scripts.
AI crawler rules1The site says which AI bots may visit, for training, search or answering questions. Blocking some is fine; what counts is making a choice.
Content Signals0The site says how its content may be used: for search, in AI answers or to train AI. Any mix of yes and no counts.
schema.org JSON-LD1Structured data on the homepage says what the site is, such as a company, a product or an article, so machines don't have to guess.
Semantic HTML1The page has a clear structure: a main content area, headings in order, a header and a footer. A bot can tell the content from the rest.
Crawlable navigation1Observed homepage navigation exposes usable page addresses and section targets in HTML. Destinations are not fetched by this check.
Named links and buttons1Links and buttons in the collected homepage HTML have names recoverable from text, labels or native alternatives, including icon controls.
Named form fields0Observed homepage fields have recoverable names and recognizable input semantics. Submission behavior is not tested.
Control roles and states0Homepage controls meet static checks for roles, required states and relationship targets, using native HTML semantics where available.
Typed links0The page links to versions made for machines, such as a Markdown copy or an API description, so an agent can go straight to them.
Canonical URL1Each page names the one address that should be indexed and cited, so an agent quoting it links to the right URL.
Working section links1Links to part of a page, such as a table of contents or a skip link, lead to that section instead of the top of the page.

Clear signals

1 site

llms-full.txt or .md pages

Websites in the directory, with their Good for Bots score. Column headers sort the list.
Site	Category	Score
Can I 3D Print It? cani3dprintit.com Decision engine providing printability verdicts, material guides, and cost estimates for objects	Research & Reference	View report for Can I 3D Print It?:

Promarks a Pro listing, in gold. Pro never changes a score or a place in the order.What Pro includes

Through a bot’s eyes

About Research & Reference

A source is useful only if a reader can tell what it actually supports. A dataset without units, a paper without a version, or a reference entry without provenance can be easy to extract and still difficult to use responsibly. For research material, the surrounding context is part of the answer.

This category brings together research publishers, data resources, encyclopedic references and other sites built to be consulted. Look for stable identifiers, accessible summaries and a clear route from an overview to the underlying material. Tables need headings and units; revisions need enough context to distinguish one edition from another.

Our scanner does not validate research or datasets. Follow the report to the source material, then check the methods, coverage and version relevant to your question.

What to look for

Use these questions when reading the sites themselves. They are guidance for this category, not additional scoring checks.

Provenance beside the claim: An entry should identify its source or method. A number without its population, time period or origin is difficult to interpret correctly.
Stable identifiers and versions: A durable link and a visible revision help a reader cite the material they actually used, especially when datasets change.
Tables with their meaning intact: Headers, units and explanatory notes should stay with the values. Extracting numbers without those labels can change the apparent result.
Reachable primary material: Summaries are useful entry points, but readers need a route to the paper, dataset or source document behind them.

Frequently asked questions

Does a high score validate a research finding?

No. It describes machine readability, not scientific validity, peer-review status or reproducibility. Those require examining the research itself.

Do you verify entire datasets?

No. The scan has a bounded request budget and examines selected public web evidence. It does not download and validate every record or supporting file.

Can restricted research still have a useful public page?

Yes. A public abstract, bibliographic record and access terms can explain what a work contains. They do not make its restricted full text available.

Questions about scores, scans and listings in general are answered in the FAQ.