Am I Ready for Agents?

Self-check for publishers and media

In short

The 18 core questions with publishing examples, plus two extra on training versus search and charging crawlers. Answer yes, not sure or no. Your answers are scored in this browser and never sent. The two extra questions sit in the find and pay areas.

Find

Can agents find you? Whether AI crawlers and assistants are allowed in, and whether they can locate your important pages.

Question 1 of 20. Has someone in your business deliberately decided which AI crawlers and assistants your robots.txt file allows, rather than leaving the default?

Example: your robots.txt treats training crawlers, AI search crawlers and assistants fetching a page for a reader as three separate decisions.

robots.txt is the first file most crawlers read, and it is where you state your position in public. It is a request rather than a lock, so it only works when it matches what your bot protection actually does.

A decision for leadershipLeadership decides, in writing, which kinds of AI visitor are welcome: training crawlers, AI search crawlers, assistants fetching a page for a person, and agents acting for a customer. The web team then makes robots.txt match that list.
The evidence behind this question (3 dated sources)
  • Ora's lab found (8 Oct 2026) that blocking agents was one of only two changes that stopped an agent getting an answer from a test site; the other was hiding the answer in JavaScript. Vendor claim. AgentReady.org guide, read 4 Nov 2026.
  • RFC 9309 states that robots.txt rules are not a form of access authorization. RFC 9309 Robots Exclusion Protocol, read 4 Nov 2026.
  • In the network's own crawl, run of 4 Nov 2026 (13,071 of 13,071 domains from Tranco list 746WX plus Ora leaderboard domains), 13.9 percent of 8,624 sites with a readable robots.txt disallowed GPTBot at the site root and 6.9 percent disallowed ChatGPT-User, the fetcher that acts on a person's question. Scorecard methodology: the crawl of 4 Nov 2026, read 4 Nov 2026.
Question 2 of 20. Has anyone checked in the last six months whether your CDN, firewall or bot protection blocks AI crawlers or AI assistants?

Example: you know whether your CDN blocks Training or Agent bots on pages with ads.

A bot rule switched on for security can turn away the assistant a customer just sent to you, and nobody in marketing sees it happen. Defaults also change: Cloudflare changed its defaults for new domains on 15 September 2026.

Take this to your web teamPlease export the current AI bot settings from our CDN or bot protection and tell us, for each of Search, Agent and Training traffic, whether it is allowed, challenged or blocked, and on which pages.
The evidence behind this question (2 dated sources)
Question 3 of 20. Do you publish an up-to-date sitemap that lists the pages you most want found?

A sitemap is the simplest list of your pages for any automated visitor. Agents mostly arrive at the home page and docs by links, so the sitemap is a backstop rather than the main route.

Take this to your web teamPlease confirm our sitemap.xml is generated automatically, includes our product, pricing, booking and policy pages, and is listed in robots.txt.
The evidence behind this question (2 dated sources)
  • Ora's traced runs found agents reached sitemap.xml in about 4 percent of runs, against 83 percent for docs pages and 69 percent for the home page (vendor research published October 2026). Vendor claim. AgentReady.org guide, read 4 Nov 2026.
  • Cloudflare's Radar scan of top domains, published 17 April 2026, found 78 percent have a robots.txt, mostly written for search engines rather than AI agents. Cloudflare: Introducing the Agent Readiness score (17 Apr 2026), read 4 Nov 2026.
Question 4 of 20. Have you decided whether to publish an llms.txt file, a short guide to your site written for language models?

llms.txt is a proposed plain-text file that points AI tools at your most useful pages. Deciding yes or not yet both count; not knowing does not.

A decision for leadershipMarketing decides whether we publish llms.txt now. If yes, the web team links it from the footer of every page, because agents reach such files by following links, not by guessing the path.
The evidence behind this question (2 dated sources)
  • Ora's lab found that 86 percent of fetches of llms.txt arrived through a link from a page the agent had already read (AgentReady guide, October 2026). Vendor claim. AgentReady.org guide, read 4 Nov 2026.
  • In the network's own crawl, run of 4 Nov 2026 (13,071 of 13,071 domains from Tranco list 746WX plus Ora leaderboard domains), 24.6 percent of 10,565 sites served an llms.txt file. Scorecard methodology: the crawl of 4 Nov 2026, read 4 Nov 2026.
Question 5 of 20. Have you decided separately whether AI systems may train on your articles and whether they may index them for search answers?

Since September 2026 some providers let you refuse training while staying in search. Treating the two as one switch costs either traffic or control.

A decision for leadershipEditorial and commercial leadership decide training and search separately, and the web team sets robots.txt and the CDN to match.
The evidence behind this question (2 dated sources)

Read

Can agents read you? Whether products, prices, policies and articles are available as text an agent can parse.

Question 6 of 20. Do your key pages (products, services, prices, policies) show their main text without needing JavaScript to run in the visitor's browser?

Many automated readers fetch the page once and do not run scripts. If your text arrives only after scripts run, those readers see an empty page and answer from somewhere else.

Take this to your web teamPlease list which of our page templates deliver their main content in the first HTML response and which need JavaScript to show it, starting with product, pricing and policy pages.
The evidence behind this question (2 dated sources)
  • In Ora's 19-configuration experiment (190 probes), hiding the answer inside a JavaScript payload stopped a plain-fetch client on 5 of 5 topics while a JavaScript-running client still passed. Vendor claim. AgentReady.org guide, read 4 Nov 2026.
  • Ora's lab found (8 Oct 2026) that answers built without reading the business's site were 3.7 times as likely to contain none of the facts the buyer asked for (25 percent versus 7 percent). Vendor claim. Ora research: AX is the new AEO (8 Oct 2026), read 4 Nov 2026.
Question 7 of 20. Are prices, availability, opening hours and delivery or cancellation terms written as text on the page, not only inside images, PDFs or pop-ups?

Example: the article text is in the HTML, with the paywall rules stated.

An agent comparing options for a customer needs the price and the terms as text it can read. If it cannot read them, it reports a gap or picks a competitor whose terms it could read.

A decision for leadershipEcommerce or product leadership decides which prices and terms are public. The web team then puts them on the page as text, including taxes, fees and cancellation rules.
The evidence behind this question (2 dated sources)
  • Ora's lab found (8 Oct 2026) that on pricing questions, answers built from the business's own pages were right 60 percent of the time against 37 percent for answers built from web search and other sites. Vendor claim. Ora research: AX is the new AEO (8 Oct 2026), read 4 Nov 2026.
  • The AgentReady guide lists plans, free tier and limits as text in the HTML, not an image, a calculator or a sales form, as a practice (AR-READ-11). AgentReady.org guide, read 4 Nov 2026.
Question 8 of 20. Do your product, event, offer or article pages carry structured data (for example Schema.org markup in JSON-LD)?

Structured data states facts such as price and availability in a form machines read directly. It helps shopping surfaces that build catalogs, though Ora's experiment found it did not change whether a plain agent retrieved the answer.

Take this to your web teamPlease tell us which Schema.org types we publish on product, offer, event and article pages, and whether the values match what the page shows.
The evidence behind this question (2 dated sources)
  • Ora's 19-configuration experiment found llms.txt, JSON-LD, sitemaps and markdown mirrors did not change whether an agent retrieved the answer; only JavaScript-hidden answers and 403 blocks did. Vendor claim. Introducing AgentReady.org (19 Oct 2026), read 4 Nov 2026.
  • In the network's own crawl, run of 4 Nov 2026 (13,071 of 13,071 domains from Tranco list 746WX plus Ora leaderboard domains), 54.0 percent of 8,783 home pages that returned 200 carried any JSON-LD. Scorecard methodology: the crawl of 4 Nov 2026, read 4 Nov 2026.
Question 9 of 20. Can your site give a clean text or markdown version of a page to an agent that asks for one?

Agents often ask for markdown because it costs them fewer tokens to read. It is a nice-to-have once the normal page works without scripts.

Take this to your web teamPlease check whether our pages answer a request with the header Accept: text/markdown with a markdown version, and tell us what it would take to add it.
The evidence behind this question (2 dated sources)

Act

Can agents act on your site? Whether the main task (buy, book, subscribe, request a quote) can be completed through standard forms or a documented interface.

Question 10 of 20. Can the main task on your site (buy, book, subscribe or request a quote) be completed with standard web forms and buttons, without a CAPTCHA at every step?

Agents working in a browser use the same forms people do. Custom widgets and repeated challenges are where they stall and hand the task back to the customer.

Take this to your web teamPlease walk our main task end to end as a guest and list every step where a CAPTCHA, a custom widget or a pop-up appears.
The evidence behind this question (2 dated sources)
  • Google said on 28 January 2026 that testers of Chrome auto browse used it for scheduling appointments, filling out online forms and getting quotes for plumbers and electricians. Google: Gemini in Chrome auto browse (28 Jan 2026), read 4 Nov 2026.
  • Ora's lab found (8 Oct 2026) that runs on sites that were not agent ready took 23 percent more turns and were blocked by the site 2.1 times as often. Vendor claim. Ora research: AX is the new AEO (8 Oct 2026), read 4 Nov 2026.
Question 11 of 20. Do your forms have clear labels and error messages that say what went wrong and how to fix it?

Clear labels help people and agents alike. An unexplained error ends the task, and the agent will not tell you it happened.

Take this to your web teamPlease check that every field in our main forms has a visible label and that each error message names the field and the fix.
The evidence behind this question (2 dated sources)
  • Ora's public check catalogue includes a usability check called Form control labeling: form fields need real labels, and placeholder text never counts. Vendor claim. Ora check catalogue API, read 4 Nov 2026.
  • Ora's lab found (8 Oct 2026) that in about 99 percent of journeys that hit a dead end on a site, the agent answered anyway from what it found elsewhere. Vendor claim. Ora research: AX is the new AEO (8 Oct 2026), read 4 Nov 2026.
Question 12 of 20. Do you offer a documented API, an MCP server or WebMCP tools for your main tasks?

These give agents a described way to act instead of clicking through pages. Assistants such as Meta's Muse use a service's API or connector when one exists and open a browser only when none does.

A decision for leadershipLeadership decides which tasks we want agents to finish directly and which stay with people. Only then does the web team pick an interface (API, MCP server or WebMCP) for those tasks.
The evidence behind this question (3 dated sources)
  • The AgentReady spec makes an MCP server or a CLI a MUST when agents should complete a task, not only read (AR-ACT-10, October 2026). AgentReady.org guide, read 4 Nov 2026.
  • In the network's own crawl, run of 4 Nov 2026 (13,071 of 13,071 domains from Tranco list 746WX plus Ora leaderboard domains), 2.1 percent of 10,385 sites published an MCP manifest, an OpenAPI file, an agent card or agents.json at the standard paths. Scorecard methodology: the crawl of 4 Nov 2026, read 4 Nov 2026.
  • Zocdoc offers registered agents an MCP server or REST API to search providers, check insurance and book appointments. Zocdoc for Agents (markdown), read 4 Nov 2026.
Question 13 of 20. Can a customer complete the main task as a guest, or sign in with a standard login flow such as OAuth rather than a custom one?

Agents acting for a person need a login flow they can recognise and a person can approve. Custom sign-in screens are a common place for an agent to stop.

Take this to your web teamPlease tell us whether our main task can be completed as a guest, and whether sign-in uses a standard flow such as OAuth with published metadata.
The evidence behind this question (2 dated sources)
  • The AgentReady guide recommends OAuth 2.0 with discoverable metadata under /.well-known (RFC 8414, RFC 9728) for user-delegated access. AgentReady.org guide, read 4 Nov 2026.
  • Cloudflare and Stripe launched a flow on 30 April 2026 in which an agent can create a Cloudflare account, start a paid plan and get an API token, with Stripe as the identity provider. Cloudflare: Agents can now create Cloudflare accounts (30 Apr 2026), read 4 Nov 2026.

Pay

Can agents pay? Whether full prices are visible before the last step, and whether your payments team has a position on purchases an agent starts.

Question 14 of 20. Is the full price, including taxes, fees and delivery, shown before the final step?

An agent comparing options for a customer needs the total, not a starting price. A total that appears only at the last step makes the comparison fail early.

A decision for leadershipEcommerce leadership decides whether the full price, fees included, is shown before checkout. The web team then makes it readable as text on the product or booking page.
The evidence behind this question (2 dated sources)
Question 15 of 20. Has your payments or fraud team agreed how to treat a purchase or booking that an AI agent starts for a customer?

Fraud tools built to stop automation may decline a legitimate agent. Someone needs to own that decision before the volume arrives.

A decision for leadershipFinance, payments and ecommerce agree a written position on agent-started purchases: allowed, allowed within limits, or handed back to the customer. Fraud rules are then set to match.
The evidence behind this question (2 dated sources)
Question 16 of 20. Has anyone in your business looked at agent payment approaches such as x402, ACP, UCP or MPP, even just to decide not yet?

Example: you have decided whether to charge AI crawlers per request, and through whom.

These are new ways for an agent to pay, or to check out on a shopping surface, without filling a card form. Knowing they exist and choosing a date to revisit is enough for this question.

A decision for leadershipThe head of ecommerce or payments names which shopping surfaces we want to sell on (Google AI Mode and Gemini, ChatGPT, Copilot, Meta) and asks our payment provider which protocol each needs.
The evidence behind this question (3 dated sources)
  • Google said on 16 September 2026 that its Universal Commerce Protocol already enables direct checkout for hundreds of thousands of brands and retailers across Google. Google: holiday shopping updates (16 Sep 2026), read 4 Nov 2026.
  • OpenAI's developer docs say building with the Agentic Commerce Protocol is open to all, while Instant Checkout in ChatGPT is available to approved partners. OpenAI commerce docs: key concepts, read 4 Nov 2026.
  • Ora's lab found agents paying through Stripe Link got 92.3 percent of test checkouts right (733 of 794 runs in a cloud browser), across 26 agent setups on the same 33 checkout tasks (AgentReady guide, October 2026). Vendor claim. AgentReady.org guide, read 4 Nov 2026.
Question 17 of 20. Have you decided whether to charge AI crawlers or agents for access, and do you know what that would cost you in reach?

Charging can bring revenue from traffic that no longer clicks through. It can also remove you from answers if the operator declines to pay.

A decision for leadershipThe commercial lead decides whether to test paid crawling on a section of the site, with a reach metric to watch.
The evidence behind this question (2 dated sources)

Trust

Can agents be trusted, and trust you? Whether you can tell verified agents from unknown automation, and whether your terms and policies are public.

Question 18 of 20. Can your team tell verified bots and agents (for example through Web Bot Auth signatures or a verified bot directory) apart from unknown automation?

Verification lets you welcome known agents without opening the door to any script that borrows their name. Without it, the choice is to let everything in or keep everything out.

Take this to your web teamPlease tell us whether our CDN or bot protection checks verified bot signatures (Web Bot Auth) or a verified bot list, and whether we treat verified agents differently from unknown automation.
The evidence behind this question (2 dated sources)
Question 19 of 20. Are your returns, cancellation, privacy and terms pages public, current and readable without logging in?

Agents and the people behind them check the rules before committing. A policy hidden in a PDF or behind a login reads as missing.

Take this to your web teamPlease confirm our returns, cancellation, privacy and terms pages load as text without JavaScript or a login, and are linked from every product or booking page.
The evidence behind this question (2 dated sources)
Question 20 of 20. Is there a named person in your business who owns how AI agents are treated on your site?

These questions cross marketing, ecommerce, security, legal and engineering. Without one owner, the bot rule, the checkout setting and the policy page drift apart.

A decision for leadershipThe leadership team names one owner for agent experience, with a quarterly review of the bot rules, readable pages, agent checkout settings and verification.
The evidence behind this question (2 dated sources)

Confirm it with a scan

Your answers show what your business has decided and checked. A scan shows what a visitor, human or agent, gets from your public pages. For leaders we recommend Ora’s scan: one score from 0 to 100 and a grade a board can track, the leaderboard of your sector to compare against, and a real agent sent through a task with its Journey test. Scanning is free and needs no account.

What it does not do: it reads public pages only, it publishes no data on repeat scans (Ora says two runs of the same scan agree within a couple of points), and its fixing service is a waitlist.

Scan your site with Ora · Which tool should our team use · What the scan tests

Your sector on Ora’s leaderboard

Your sector on Ora's leaderboard: scores of the 110 listed publishing sitesBAmedian 4701020273040402350166027028090
Your sector on Ora's leaderboard: scores of the 110 listed publishing sites. Number of sites per score band: 27 in 30 to 39, 40 in 40 to 49, 23 in 50 to 59, 16 in 60 to 69, 2 in 70 to 79, 2 in 80 to 89. The median is 47. A B starts at 70 and an A at 86. Ora's leaderboard, 5 Nov 2026.

On 5 Nov 2026 Ora listed 110 sites in Gaming & Media and News & Intelligence. The median was 47 and the top score 89; the highest three were magichour.ai 89, beehiiv.com 88 and ZoomInfo 78. Scores come from scanning sites, not from answers, so scan yours and find its place on this chart.

About this version

The two extra questions are marked in the list and count inside their own area, so this version scores out of 20 questions. The scoring rule is the same as the core self-check. Read the publishing guide for the dated examples behind the extra questions.

Extra questions: Have you decided separately whether AI systems may train on your articles and whether they may index them for search answers? Have you decided whether to charge AI crawlers or agents for access, and do you know what that would cost you in reach?

Sources

  1. AgentReady.org guide (agentready.org), read 4 Nov 2026. Vendor claim.
  2. RFC 9309 Robots Exclusion Protocol (rfc-editor.org/rfc/rfc9309), read 4 Nov 2026.
  3. Scorecard methodology: the crawl of 4 Nov 2026 (amireadyforagents.com/scorecards/methodology), read 4 Nov 2026.
  4. Cloudflare changelog: New options to manage AI traffic (1 Jul 2026) (developers.cloudflare.com/changelog/post/2026-07-01-ai-traffic-options), read 4 Nov 2026.
  5. Cloudflare: Introducing the Agent Readiness score (17 Apr 2026) (blog.cloudflare.com/agent-readiness), read 4 Nov 2026.
  6. Cloudflare: Disallow AI Training, Accountable crawlers (15 Sep 2026) (blog.cloudflare.com/accountable-mixed-use-ai-crawlers), read 4 Nov 2026.
  7. Cloudflare: The Internet has a second audience (30 Sep 2026) (blog.cloudflare.com/agentic-web), read 4 Nov 2026.
  8. Ora research: AX is the new AEO (8 Oct 2026) (ora.ai/blog/ax-is-the-new-aeo), read 4 Nov 2026. Vendor claim.
  9. Introducing AgentReady.org (19 Oct 2026) (ora.ai/blog/agentready-launch), read 4 Nov 2026. Vendor claim.
  10. Google: Gemini in Chrome auto browse (28 Jan 2026) (blog.google/products-and-platforms/products/chrome/gemini-3-auto-browse), read 4 Nov 2026.
  11. Ora check catalogue API (ora.ai/api/checks), read 4 Nov 2026. Vendor claim.
  12. Zocdoc for Agents (markdown) (zocdoc.com/business/welcome-agents.md), read 4 Nov 2026.
  13. Cloudflare: Agents can now create Cloudflare accounts (30 Apr 2026) (blog.cloudflare.com/agents-stripe-projects), read 4 Nov 2026.
  14. Google: How to book travel with AI Mode (27 Aug 2026) (blog.google/products-and-platforms/products/search/book-travel-ai-mode), read 4 Nov 2026.
  15. Retail Gazette: Amazon blocks Meta's Muse (22 Sep 2026) (retailgazette.co.uk/blog/2026/09/amazon-blocks-metas-new-ai-shopping-agent-from-its-website), read 4 Nov 2026. Reported by trade press.
  16. Shopify Help Center: Meta agentic storefront (help.shopify.com/en/manual/online-sales-channels/agentic-storefronts/meta), read 4 Nov 2026.
  17. Google: holiday shopping updates (16 Sep 2026) (blog.google/products-and-platforms/products/shopping/google-shopping-updates-holiday-shopping), read 4 Nov 2026.
  18. OpenAI commerce docs: key concepts (developers.openai.com/commerce/guides/key-concepts), read 4 Nov 2026.
  19. Cloudflare: the agentic internet, WebMCP example (6 Aug 2026) (blog.cloudflare.com/the-agentic-internet), read 4 Nov 2026.
  20. Ora leaderboard API (GET /api/leaderboard) (ora.ai/api/leaderboard), read 5 Nov 2026. Vendor claim.
  21. Ora public leaderboard, category averages (ora.ai/leaderboard), read 5 Nov 2026. Vendor claim.
  22. Ora pricing (JSON) (ora.ai/pricing), read 5 Nov 2026. Vendor claim.
  23. Ora llms.txt (ora.ai/llms.txt), read 5 Nov 2026. Vendor claim.
  24. Ora home page (ora.ai), read 5 Nov 2026. Vendor claim.
  25. Ora: Introducing Deep Scan (8 Apr 2026) (ora.ai/blog/introducing-deep-scan), read 5 Nov 2026. Vendor claim.
  26. Sector scorecards dataset, edition 4 Nov 2026 (amireadyforagents.com/scorecards), read 4 Nov 2026.

Last reviewed