An Agentic Score is a 0 to 100 measure of how ready a website is for AI agents: whether an agent can find the site, read its pages, and act on what it finds. It comes from Is Agentic, a free agent-readiness tool from Vercel, with scans run by Ora, and it is scored against AgentReady.org, an open standard developed with Ora, Vercel, Stripe, Google, and others. Connectively now shows it next to Domain Authority and Domain Rating for the publications behind 98.6% of open questions, so experts can check a publication the way journalists check an expert.
This article covers what the score measures, what Connectively's own data shows, how the score is calculated, how it differs from the other agent-readiness scores you will run into, what the evidence says about publishers and AI citations, and how to use the score when you decide which questions to answer.
What Connectively's data shows
Connectively compared Agentic Score with Domain Authority and Domain Rating across 906 publications that post questions to experts. Domain Authority does not predict agent readiness. The correlation between the two is 0.08. Domain Rating does no better, at 0.10. Domain Authority and Domain Rating correlate with each other at 0.87, so the two authority metrics measure the same thing, and neither measures this.
| Domain Authority band | Publications | Average Agentic Score | Score 70 or higher | Score below 40 |
|---|---|---|---|---|
| DA 0 to 19 | 355 | 36.9 | 2.5% | 64.2% |
| DA 20 to 39 | 265 | 39.7 | 3.4% | 55.5% |
| DA 40 to 69 | 177 | 42.2 | 11.3% | 50.8% |
| DA 70 to 100 | 110 | 39.9 | 8.2% | 51.8% |
Source: Connectively Agent Readiness Report, November 2, 2026. Excludes 44 publications owned by Featured, which are scored on the same basis but were removed so that our own sites do not shape the result. One publication has no Domain Authority.
Publications with a Domain Authority above 70 average 39.9. Publications below 20 average 36.9. The spread between the top and bottom bands is three points on a 100-point scale. Of the 143 publications with a Domain Authority of 60 or higher, 79 (55.2%) score below 40. Across all 908 scored publications, the average is 39.1, the median is 37, 57.5% score below 40, and 5.2% score 70 or higher. That is close to the web as a whole: Ora reports an average of 43 across 127,521 scanned sites, with 59% at a D or F.
How the score is calculated
Is Agentic publishes its methodology. Checks fall into three groups:
| Group | Points | What it covers |
|---|---|---|
| Essential | 80 | Not blocking agents, content available without JavaScript or sign-in, clean redirects, agent-readable 404s, an OpenAPI spec, JSON error responses, Markdown content negotiation |
| Recommended | 20 | Sitemap, JSON-LD structured data, metadata completeness, organization schema, pricing and trust pages, API documentation linked from the homepage |
| Bonus | Up to 5 | Emerging signals such as llms.txt, sitemap freshness, developer portals, and accessibility structure |
Checks that do not apply to a site type are excluded, partial results receive proportional credit, and API or MCP checks activate only when that surface exists. Ora caches completed scans for six hours, and Is Agentic stores the latest result as a public report with a timestamp. Connectively re-pulls each publication's score every 15 days.
Two caveats. First, Is Agentic describes a score as a snapshot and does not promise that an agent can use a site. Second, the checks are not equally predictive. AgentReady's own tests found that two changes broke an agent: hiding the answer behind JavaScript, and blocking access. Content negotiation and .md endpoints "did not change whether agents succeeded," yet Markdown negotiation is an Essential check. A site that agents can read can still score low on it.
Scores you will confuse with this one
Three other tools score agent readiness, and each answers a different question.
- Lighthouse's Agentic Browsing category. Shipped in Lighthouse 13.3 in May 2026 and still marked "under development" in Chrome's docs. It checks llms.txt, WebMCP, the accessibility tree, and cumulative layout shift, and reports a category score out of 100. It runs on one page at a time inside Chrome DevTools, and it is aimed at developers testing their own pages.
- Cloudflare's Agent Readiness score. Scans the 200,000 most visited domains for robots.txt and sitemap discoverability, Markdown for agents, bot access rules including Content Signals, and capability discovery such as API catalogs and MCP server cards. Cloudflare found Markdown content negotiation passes on 3.9% of sites.
- Mintlify's Agent Score. Runs 29 checks on documentation sites against the Agent-Friendly Documentation Spec, covering discoverability, Markdown availability, page size, structure, URL stability, observability, and authentication. It is built for docs, not publications.
The Agentic Score Connectively shows is the Is Agentic score: a whole-site scan against the AgentReady standard, with a public report per domain. It is the one of the four that is run against publications at scale and refreshed on a schedule, which is why Connectively uses it.
What the evidence says about publishers
News publishers block AI agents more than most sites. BuzzStream's December 2025 audit of the top 50 US and UK news sites found 79% block at least one AI training bot and 62% block GPTBot. The Reuters Institute found 79% of the most-used US news sites blocked OpenAI's crawlers by the end of 2023.
Blocking has consequences, but only with the agents that honor it. Goodie's July 2026 study of 31 million AI citations found that The New York Times, which blocks all 13 major AI user agents, received zero citations from ChatGPT and Gemini and one from Claude. The block did not stop Grok (37,642 citations), Google AI Overviews, or Perplexity.
The evidence also cuts the other way, and it is worth knowing. In the same Goodie study, Forbes holds 33% of all news citations while scoring 45% on Goodie's readiness audit. Readiness is not the only thing that decides citations. Authority and reach still matter a great deal. The Featured AI Citation Report, which analyzed 22,881 citations returned by Perplexity across 405 GEO audits between June 2 and August 21, 2026, found that 34.5% of citations pointed to domains with a Domain Authority under 40, and that 31.2% pointed to domains at DA 80 and above. Both tails are large. The middle is thin.
Other studies fit that picture. Ahrefs found that 37.9% of the 4 million URLs cited in Google AI Overviews ranked in the top 10 for the query, 31.2% ranked 11 to 100, and 31.0% did not rank in the top 100 at all. Muck Rack's Generative Pulse, which analyzed more than 25 million links cited by ChatGPT, Claude, and Gemini, found that earned media accounts for 84% of AI citations.
What this means for a pitch list
Domain Authority tells you how a publication ranks in search. The Agentic Score tells you how ready its site is for the agents that answer questions in ChatGPT, Perplexity, Claude, and Gemini. The two are unrelated, so a list sorted by one tells you nothing about the other.
- Look at both before you answer. On Connectively, turn on the Agentic Score and Domain Rating columns and sort by whichever matters for the quote you are writing. A question from a publication with a high Domain Authority and a low Agentic Score is still worth answering for readers and backlinks. A question from a publication with a high Agentic Score sits on a site that agents can reach.
- Do not screen on Domain Authority alone. Connectively's data shows the same share of agent-ready publications in the bottom DA bands as in the top ones. A DA filter removes them for no reason.
- Check your own site. Run your company domain through is-agentic.com. The two fixes that matter most, according to AgentReady's tests, are letting agents in and putting the answer in the initial HTML rather than behind JavaScript.
- Use the refresh. Scores change every 15 days as publications change their crawler rules. Re-check rather than keeping a static tier sheet.
How Connectively shows the score
Every Connectively table that shows a publication's Domain Authority can also show its Domain Rating and Agentic Score. On the Questions page, click the sliders button at the right end of the toolbar and tick the two columns. They appear next to Domain Authority, sort on a click, and show their source on hover. The score also appears on each publication's public page. Agentic Score is on every plan, including Free. Domain Rating is included on [PLANS: confirm with Adrian]. Both are returned by the Connectively API. The step-by-step guide is on the Connectively blog.
Journalists use Connectively's Verification Signals to check an expert. Experts can now check a publication the same way.
Frequently asked questions
What is a good Agentic Score?
Is Agentic does not publish grade bands. In Connectively's data, a score of 70 or higher means the site passes most Essential checks, and only 5.2% of publications reach it. A score below 40 means the site fails most Essential checks, which cover blocking agents, content behind JavaScript or a login, redirects, error pages, and API documentation.
Does a low Agentic Score mean AI will not cite the publication?
No. Forbes scored 45% on Goodie's readiness audit and holds a third of all news citations. The score measures readiness, not citations.
Does a high Agentic Score guarantee citations?
No. It means agents can reach and retrieve the site. Relevance, freshness, and how often other sources mention a brand all matter.
How is an Agentic Score different from Domain Rating?
Domain Rating, from Ahrefs, measures the strength of a site's backlink profile. The Agentic Score measures agent readiness. Across 906 publications on Connectively the two correlate at 0.10.
How often does the score change?
Connectively refreshes it every 15 days. Is Agentic re-scans when a report is requested and the cached result is older than six hours.
