Calstart Inc
12 of 38 tests pass 43 points still on the table
Cohort median 54, range 26–76. Grey is the middle half.
The fundamentals are in place and the machine-readable layer on top of them largely is not.
The fastest 14 points on this scan
Does your website let those crawlers in, or does your security turn them away?
1 of 12 search crawlers and user-triggered fetchers are rejected or challenged at the edge. 1 of 28 sampled pages return an interactive challenge (edge challenge cookie).
Do this: Stop rejecting AI crawlers at the CDN or WAF.
Is the text behind tabs and 'read more' buttons already on the page?
0 of 1 pages ship every tab and 'read more' panel in the page itself and keep it there. 1 remove panel text with a script until it is clicked.
Do this: Render every panel and hide it with CSS.
Do your buttons, links and menus have names an AI agent can read?
4 of 28 sampled pages give nearly every button, link and menu a name; 53 of 3249 controls have none.
Do this: Give every button, link and menu a name.
Where Calstart Inc’s 57 came from
Every test is listed below, grouped the same way, worst first.
| Dimension | Share earned | Earned | Score |
|---|---|---|---|
| 1 · Reach | 18 | 72 | |
| 2 · Read | 17 | 49 | |
| 3 · Understand | 8 | 32 | |
| 4 · Act | 9 | 60 |
The work queue
38 scored tests
RCH-1 Can the AI tools that search the web read your site? PASS 8 pt
All 9 search crawlers may fetch content paths.
Your robots.txt tells crawlers what they may read. If it shuts out the search crawler behind an assistant, that assistant cannot cite you.
RCH-2 Does your website let those crawlers in, or does your security turn them away?1 of 12 search crawlers and user-triggered fetchers are rejected or challenged at the edge. 1 of 28 sampled pages return an interactive challenge (edge challenge cookie). NEEDS WORK +6
1 of 12 search crawlers and user-triggered fetchers are rejected or challenged at the edge. 1 of 28 sampled pages return an interactive challenge (edge challenge cookie).
A firewall or bot rule can turn a crawler away even when robots.txt says yes, or put a challenge page in front of it that it cannot pass. We request pages the way each crawler does and record what comes back.
Sources (3)
- Cloudflare, 'Perplexity is using stealth, undeclared crawlers' A
- Cloudflare, new AI traffic options A
- RESONEO, LLM crawler test B
- Operations (lead)
- Back-end developer
RCH-3 Does each page say which web address is the real one?11% of sampled pages emit a valid, self-consistent canonical. 25 have problems. PROBLEM +1
11% of sampled pages emit a valid, self-consistent canonical. 25 have problems.
A canonical link names the true address of a page. Without one, the same content can read as several competing pages.
Sources (2)
- Back-end developer (lead)
- Search specialist
RCH-4 Do your page settings let assistants quote you? PASS 3 pt
No noindex, nosnippet, max-snippet:0, noarchive or nocache directive found on the sampled pages.
Page tags and headers that stop a search engine from showing a snippet, or from indexing a page at all, also keep that page out of the AI answers built on the search index.
RED-1 Is your page's text in the page itself, without needing JavaScript? PASS 10 pt
The first response from your server holds 99% of the text a browser shows, on average across 28 pages; 1% of the words are missing until scripts run. The worst template is /: 87% across 2 pages.
If your text only appears after JavaScript runs, a tool that does not run it sees an empty page. This is the heaviest test in the Index.
RED-2 Is the text behind tabs and 'read more' buttons already on the page?0 of 1 pages ship every tab and 'read more' panel in the page itself and keep it there. 1 remove panel text with a script until it is clicked. PROBLEM +4
0 of 1 pages ship every tab and 'read more' panel in the page itself and keep it there. 1 remove panel text with a script until it is clicked.
Tabs and accordions are fine, as long as their text is already in the page before anybody clicks.
Sources (3)
- Front-end developer
RED-3 Can a crawler trust the 'last changed' dates in your sitemap?No sitemap URLs with lastmod values were found. PROBLEM +2
No sitemap URLs with lastmod values were found.
A last-changed date should mean the content changed, not that you deployed. Once it is noise, crawlers stop trusting it.
- Back-end developer (lead)
- Search specialist
RED-4 Do your filters and calendars keep a crawler from getting lost? PASS 1 pt
No parameterized URLs appeared in the crawlable sample.
Filters and calendars can generate millions of addresses. A crawler spends its visit there instead of on the pages you care about.
Sources (1)
RED-5 Do your PDFs and images have their own rules for search and quoting?0 of 25 non-HTML resources carry an X-Robots-Tag. PROBLEM +1
0 of 25 non-HTML resources carry an X-Robots-Tag.
PDFs, images and JSON cannot carry a meta robots tag, so the instruction has to travel in the HTTP header instead.
Sources (1)
- Operations
RED-6 Do missing pages say they are missing? PASS 1 pt
0 of 10 deliberately invalid URLs returned HTTP 200.
A missing page that answers 200 OK looks like a real page, and a crawler spends its visit on it.
RED-7 Do your redirects go straight to the right page? PASS 1 pt
27 of 28 sampled URLs redirect.
Every extra redirect is another chance for a crawler to give up. One hop, straight to the destination.
Sources (1)
RED-8 Do you have sitemaps for your images and videos?No sitemap could be fetched or parsed. PROBLEM +1
No sitemap could be fetched or parsed.
Image and video sitemaps tell a crawler about media it would not find in the page text.
Sources (1)
- Back-end developer (lead)
- Search specialist
RED-10 Do you offer a plain-text (Markdown) copy of your pages?0 of 24 sampled pages have a Markdown copy that matches the page and carries no navigation (looked for at the page address plus .md; no page advertised one). PROBLEM +2
0 of 24 sampled pages have a Markdown copy that matches the page and carries no navigation (looked for at the page address plus .md; no page advertised one).
A plain-text Markdown copy of a page has none of the navigation and markup a tool has to wade through to reach your words.
- Back-end developer
RED-12 If a machine asks for Markdown, does your site give it?0 of 25 URLs serve Markdown for Accept: text/markdown. PROBLEM +2
0 of 25 URLs serve Markdown for Accept: text/markdown.
A machine can ask the page's own address for Markdown and get it, with caches keeping the web and Markdown versions apart.
- Back-end developer (lead)
- Operations
RED-13 Is the page built from real page parts, like header, main and navigation?21% of sampled pages carry one main element, a wrapping article or section, and a nav. 22 exceed 12 divs per semantic element. PROBLEM +1
21% of sampled pages carry one main element, a wrapping article or section, and a nav. 22 exceed 12 divs per semantic element.
Real page parts such as header, main and navigation, instead of one pile of unlabeled boxes. It is good practice for people, and research suggests structure helps a model read a page.
Sources (2)
- Front-end developer
RED-14 Do your headings follow in order, starting from one main heading?43% of sampled pages have exactly one non-empty h1 and no skipped heading levels. 9 have no h1; 1 have several; 7 skip a level. NEEDS WORK +0
43% of sampled pages have exactly one non-empty h1 and no skipped heading levels. 9 have no h1; 1 have several; 7 skip a level.
One main heading, then sub-headings in order. A study found most pages ChatGPT cited used ordered headings; that shows association, not cause.
Sources (2)
- Front-end developer (lead)
- Content team
RED-16 Does every section heading have a link that will not change?0 of 1079 subheadings carry a slug-like id (0%); 0 have generated or unstable ids. PROBLEM +1
0 of 1079 subheadings carry a slug-like id (0%); 0 have generated or unstable ids.
A section heading with a stable ID can be linked to directly, so a person or a tool can return to the same passage.
Sources (1)
- Front-end developer
RED-17 Do your images have descriptions for people who cannot see them?40 of 244 images carry appropriate alt text: 63 have no alt attribute, 4 use the filename. PROBLEM +1
40 of 244 images carry appropriate alt text: 63 have no alt attribute, 4 use the filename.
Alt text is the only text an image has. Decorative images should say so with an empty alt rather than describing themselves.
Sources (1)
- Content team (lead)
- Back-end developer
RED-18 Are your key facts written as text, not only inside images or PDFs?6 of 28 sampled pages are thin relative to their images or lead with a document download. NEEDS WORK +0
6 of 28 sampled pages are thin relative to their images or lead with a document download.
A fact that exists only in a PDF, a scan or an infographic is harder to find and quote. It is not a hard barrier: at least one AI tool reads PDFs.
Sources (1)
- Content team (lead)
- Front-end developer
RED-19 Does each page give its answer near the top? PASS 1 pt
79% of sampled pages open with a full sentence of at least 8 words within their first 500 characters.
A page that states its answer in the first sentences is easier to quote. Bing's guidelines advise it, and one study found 44% of ChatGPT citations came from the first 30% of a page.
Sources (2)
- Bing Webmaster Guidelines A
- a study of which part of a page ChatGPT cites B
UND-1 Does your site say clearly who you are?0 of 7 organizational facts are present in the Organization markup: none. PROBLEM +2
0 of 7 organizational facts are present in the Organization markup: none.
Your name, address, contact details and public identifiers, on an About page and in your page code, so a machine can tell which organization you are.
Sources (1)
- Content team (lead)
- Site owner
UND-2 Does your site use the page labels that search engines still read for what you publish?0 of 4 expected nonprofit types are present: none. Missing: NGO, DonateAction, Grant, FundingScheme. PROBLEM +3
0 of 4 expected nonprofit types are present: none. Missing: NGO, DonateAction, Grant, FundingScheme.
Use the markup types that match what you publish, such as local business, jobs or courses, and that search engines still read.
- Back-end developer (lead)
- Search specialist
UND-4 Does each article show who wrote it, with a page about the author? PASS 2 pt
3 of 3 distinct authors have a page or profile a machine can find (an author url, a sameAs, or a link on the byline).
An article should show who wrote it, and the name should lead to a page or profile about that person.
UND-5 Does the author label in your page code point to that person?0 of 27 substantive pages show a byline and point to the author with a reference, a link or a profile; 0 carry a bare string; 27 are unsigned. PROBLEM +1
0 of 27 substantive pages show a byline and point to the author with a reference, a link or a profile; 0 carry a bare string; 27 are unsigned.
The author label in your page code should point to the person, not repeat their name as text on every article.
Sources (1)
- Back-end developer (lead)
- Content team
UND-6 Are your published and updated dates honest and visible?27 of 28 pages carry a publication or modification date in their markup, 4 of them show a date to the reader, across 22 distinct dateModified values. 4 are inconsistent. NEEDS WORK +3
27 of 28 pages carry a publication or modification date in their markup, 4 of them show a date to the reader, across 22 distinct dateModified values. 4 are inconsistent.
Published and updated dates should be visible to readers, match your page code, and move only when the content does. A date that changes on every deploy is noise.
- Back-end developer (lead)
- Content team
UND-7 Does each page show where it sits in your site? PASS 1 pt
The sampled pages are all at the first level of the site, so there is no hierarchy for a trail to show.
Where your site has sections, each page should show where it sits, so a model knows a program page belongs to a department.
Sources (1)
UND-8 Does your hidden page code say the same thing as the visible text?On 0% of the 1 sampled pages with markup, the headline or name in the page code appears in the visible page. PROBLEM +1
On 0% of the 1 sampled pages with markup, the headline or name in the page code appears in the visible page.
The headline and name in your page code should be the ones on the page. Markup that says something else is ignored at best, and misleading at worst.
Sources (1)
- Back-end developer (lead)
- Front-end developer
UND-10 Does your page code link to your official profiles elsewhere?No sameAs links were found on Organization or Person nodes. PROBLEM +2
No sameAs links were found on Organization or Person nodes.
Links from your organization's page code to your official profiles, such as Wikidata and LinkedIn.
Sources (1)
- Search specialist (lead)
- Back-end developer
UND-11 Does your page code pass a validity check? PASS 2 pt
260 nodes across 28 pages parse, carry what their types require, and hold well-formed values.
Markup that does not parse, or that leaves out what its type needs, is served to machines that cannot use it, and nothing on the page says so.
Sources (1)
ACT-1 Can an AI agent tell what each form field is for? PASS 5 pt
1 of 1 form fields carry both an accessible name and a meaningful name attribute.
An agent filling in your form has to know what each field is for. Labels and names are how it finds out.
ACT-2 Can an AI agent browse and search without hitting a CAPTCHA? PASS 3 pt
No search or filter surface in the sample returned a challenge.
Challenge the actions that need challenging. Browsing, searching and filtering are not those actions, and a CAPTCHA there ends an agent's visit.
Sources (2)
- OpenAI, Operator (hands off captchas) B
- Open CaptchaWorld (benchmark paper) B
ACT-4 Do your buttons, links and menus have names an AI agent can read?4 of 28 sampled pages give nearly every button, link and menu a name; 53 of 3249 controls have none. PROBLEM +4
4 of 28 sampled pages give nearly every button, link and menu a name; 53 of 3249 controls have none.
An agent that clicks through your site finds each control by its name, and a control with no name is one it cannot choose.
Sources (2)
- Anthropic, browser use tool (accessibility tree) A
- OpenAI's guidance on ARIA labels for ChatGPT Atlas, reported by Adrian Roselli (address to confirm: OpenAI's own page was not opened) C
- Front-end developer (lead)
- Content team
PST-1 Which AI training crawlers have you allowed or blocked? NOT SCORED —
Training crawlers and opt-out tokens blocked: none. Allowed: GPTBot, ClaudeBot, CCBot, Bytespider, Google-Extended, Applebot-Extended. ChatGPT-User, Claude-User, Perplexity-User act on a person's request and may ignore robots.txt. robots.txt does not name any of xAI's observed crawlers.
Which AI training crawlers you allow or block is your choice, so it is shown and never scored.
Sources (7)
- OpenAI, crawler overview A
- Anthropic Help Center, crawler controls A
- Google, common crawlers (Google-Extended) A
- Apple, About Applebot A
- Google, user-triggered fetchers A
- Known Agents and Botcrawl directories (xAI crawler names; third-party, xAI publishes none) C
- StackFox, test of Grok's user agent (single blog, unverified) C
PST-2 Have you posted a Content Signals statement about what AI may do with your content?No Content-Signal line in robots.txt. Missing: search, ai-input, ai-train. NOT SCORED —
No Content-Signal line in robots.txt. Missing: search, ai-input, ai-train.
A line in robots.txt or in a response header that states whether AI may search, quote or train on your content. Having a stated position is the point.
- Search specialist (lead)
- Site owner
PST-4 Have you reserved your text and data mining rights in a form machines can read?No TDMRep reservation was found. NOT SCORED —
No TDMRep reservation was found.
A reservation of your text and data mining rights, in a form machines can read. TDMRep is a W3C community standard tied to EU copyright law.
Sources (1)
- Site owner (lead)
- Back-end developer
PST-5 Do your statements about AI use agree with each other?No statement about AI use was found, so there is nothing to compare. NOT SCORED —
No statement about AI use was found, so there is nothing to compare.
Your statements about AI use should say the same thing. A Markdown copy that allows AI training while robots.txt forbids it tells a crawler two things.
Sources (1)
- Site owner (lead)
- Search specialist
PST-6 Is your site switched on for Google's AI answers?Google lets owners switch a site out of AI Overviews and AI Mode in Search Console. The setting cannot be seen from outside. NOT SCORED —
Google lets owners switch a site out of AI Overviews and AI Mode in Search Console. The setting cannot be seen from outside.
Google lets site owners switch their site out of its AI answers. A scan cannot see the setting, so we ask you to confirm it.
Sources (1)
- Search specialist (lead)
- Site owner
EXT-1 Do you publish an llms.txt file for coding assistants?/llms.txt returns HTTP 404. NOT SCORED —
/llms.txt returns HTTP 404.
A file at /llms.txt listing your most important pages, for coding assistants and documentation tools. Search crawlers do not read it.
- Back-end developer (lead)
- Search specialist
EXT-4 Can an AI assistant connect to your site as a tool?No domain MCP server or tool descriptor was discoverable. NOT SCORED —
No domain MCP server or tool descriptor was discoverable.
A public MCP server over your catalog or directory, so an assistant can be pointed at it and call it as a tool rather than scraping a results page.
Sources (1)
- Back-end developer (lead)
- Operations
EXT-5 Is your page code one connected description of your organization?29 of 66 entity nodes carry an @id (44%). The WebPage to WebSite to Organization chain is broken at WebSite->Organization. NOT SCORED —
29 of 66 entity nodes carry an @id (44%). The WebPage to WebSite to Organization chain is broken at WebSite->Organization.
Structured data that links together, your organization, your site and this page, instead of a scattering of disconnected labels.
Sources (2)
- Back-end developer (lead)
- Search specialist
EXT-6 Is your page code generated from your content rather than typed by hand?Schema values track visible content across 0 templates. NOT SCORED —
Schema values track visible content across 0 templates.
Structured data generated from your content model survives the next content change. Hardcoded markup rots quietly.
Sources (1)
- Back-end developer
EXT-11 Have you marked the part of a page that answers a spoken question?No speakable specification found on any sampled page. NOT SCORED —
No speakable specification found on any sampled page.
Flag the part of a page that answers a spoken question, the lede or the summary, so an assistant reads that rather than guessing at a paragraph.
Sources (1)
- Back-end developer (lead)
- Content team
EXT-12 Do you tell search engines the moment a page changes?No IndexNow key file was discoverable. The file is named after the key, which is not published, so this is 'not discoverable' rather than proof of absence. NOT SCORED —
No IndexNow key file was discoverable. The file is named after the key, which is not published, so this is 'not discoverable' rather than proof of absence.
Tell search engines the moment a page changes, rather than waiting to be crawled again.
Sources (1)
- Back-end developer (lead)
- Search specialist
EXT-13 Can an assistant ask your site a question directly?No NLWeb endpoint was discoverable. NOT SCORED —
No NLWeb endpoint was discoverable.
A natural-language endpoint over your own structured data, so an assistant can ask your site a question instead of scraping around it.
Sources (1)
- NLWeb A
- Back-end developer (lead)
- Operations
EXT-14 Does your site offer its actions to AI browsers as tools?No WebMCP tool was registered or declared. NOT SCORED —
No WebMCP tool was registered or declared.
A page can hand a browser agent its actions by name instead of leaving the agent to work out its buttons.
- Front-end developer (lead)
- Back-end developer
EXT-16 Do you publish a card or a skills list for the agents you run?No agent card or skills index was found. NOT SCORED —
No agent card or skills index was found.
If your site runs an agent, a card tells other agents what it can do and where to reach it.
- Back-end developer (lead)
- Operations
EXT-17 Do you list your agents in your domain's DNS records?No SVCB record at _index._agents.calstart.org. NOT SCORED —
No SVCB record at _index._agents.calstart.org.
A domain can announce its agents in DNS, where an agent looks before it fetches anything.
Sources (1)
- Operations
EXT-19 Does robots.txt say where your sitemap is?No sitemap was found, and robots.txt names none. NOT SCORED —
No sitemap was found, and robots.txt names none.
A sitemap that robots.txt names is found by every crawler without guessing an address.
Sources (2)
- Search specialist (lead)
- Back-end developer
EXT-20 Can an AI browser see the controls on your page? NOT SCORED —
28 of 28 sampled pages have no control hidden from the accessibility tree, no invalid role and no role without its parent; 0 problems in all.
A browser agent finds what it can press through the accessibility tree, and a broken tree hides or mislabels controls from it.
MSR-2 Can you count visits that come from AI assistants?Whether a channel group exists for AI assistant referrers is not visible from outside. Analytics presence is; the grouping is not. NOT SCORED —
Whether a channel group exists for AI assistant referrers is not visible from outside. Analytics presence is; the grouping is not.
Visits from ChatGPT, Perplexity, Claude and the rest land in Direct by default, where they disappear. A channel group makes them countable.
Sources (2)
- Google Analytics, default channel groups A
- Google Analytics AI Assistant channel B
- Search specialist
MSR-3 Do you keep your crawler logs long enough to use them?Log retention and queryability are not observable from outside the site. NOT SCORED —
Log retention and queryability are not observable from outside the site.
Crawler logs have to survive long enough to show a before and an after. Many hosts discard them in days.
Sources (1)
- Operations
MSR-5 Do you ask AI assistants the same questions every month and keep the answers?Answer-share tracking happens outside the site and must be attested. NOT SCORED —
Answer-share tracking happens outside the site and must be attested.
Ask the same questions of the same assistants every month and keep the answers, so you can see whether what they say about you is changing.
Sources (1)
- Search specialist
MSR-6 Have you measured how accurately an AI reads your page?No extraction-fidelity baseline is captured by this version of the scanner, which makes no live model calls. NOT SCORED —
No extraction-fidelity baseline is captured by this version of the scanner, which makes no live model calls.
Feed a page to an assistant, ask it your key questions, and score the answers. Do it again after the fixes and compare.
Sources (1)
- Search specialist (lead)
- Content team
MSR-7 Are you set up in Google Search Console and Bing Webmaster Tools, and do you read their AI reports?Google Search Console verification found on the page. Whether sitemaps are submitted and alerts reach a person cannot be seen from outside. NOT SCORED —
Google Search Console verification found on the page. Whether sitemaps are submitted and alerts reach a person cannot be seen from outside.
Verify your site in both Google Search Console and Bing Webmaster Tools, and read their AI reports. Bing's index feeds several AI products.
Sources (1)
- Search specialist (lead)
- Operations
MSR-8 Does your page hold still while it loads? NOT SCORED —
28 of 28 rendered pages shifted by 0.1 or less; the worst shifted 0.06.
An agent that acts from a screenshot clicks where a button was, so a page that moves under it makes it miss.
28 pages, sampled from this site’s own sitemap and internal links
The cap is 32, a ceiling and not a certainty: a site with fewer reachable pages yields fewer.
28 pages crawled
- www.calstart.org
- calstart.org/home
- calstart.org/about
- calstart.org/contact
- calstart.org/mission
- calstart.org/national-influence
- calstart.org/global-influence
- calstart.org/projects
- calstart.org/board-members
- calstart.org/leadership
- calstart.org/initiatives
- calstart.org/light-duty
- calstart.org/buses
- calstart.org/trucks
- calstart.org/infrastructure
- calstart.org/innovative-mobility
- calstart.org/activities
- calstart.org/technology-development-and-demonstration
- calstart.org/public-policy
- calstart.org/market-acceleration
- calstart.org/dna
- calstart.org/membership
- calstart.org/resources
- calstart.org/events
- calstart.org/calstart-news
- calstart.org/press-releases
- calstart.org/blog
- calstart.org
28 other pages, all 200, by template
Everything needed to check this number
| Cohort | nonprofits |
|---|---|
| Source | Exempt Organizations Business Master File Extract, with IRS Form 990 filings (Internal Revenue Service (IRS), with Wikidata, 2026-09-07) |
| Index | 12.0, in effect 2026-10-18 |
| Scanner | 0.1.0 |
| Scanned | 2026-10-19T19:15:47.018566+00:00 |
| Result hash | e71960fd35001e63b9fcb16b05ec6609cfcddf572fb5adcb3b9860c4053724ef |
| Basis | public-interest census: a declared bot with a published policy at https://readinessindex.io/bot, honoring robots.txt addressed to AIRBot, reading public pages only. No per-site permission was sought or given. |
The hash is the SHA-256 of the full result.json,
recorded when the scan finished rather than computed now. The raw evidence
behind every line above, every response, cached, is retained and is a
median 1.5 MB per site, so it is not served from here.
Ask us for it if you want to
check a number against the bytes it came from.
Think this is wrong? It might be. Tell us at hello@readinessindex.io and we will look, correct it if it is, and say so publicly either way.
About these numbers. Points are whole numbers everywhere on this page. The arithmetic behind them is not: a test scores its weight times its band divided by four, so a row can earn 1.5 points and print as 2. Scores are computed from the exact figures and rounded once, at the end, never from the rounded rows. The rule is published in the standard.