Three perfect pillars, one near-zero. Here's exactly where our site was ready for AI agents, where it wasn't, and what the gap tells you about the web everyone is about to have to fix.
There's an uncomfortable moment that comes with building a tool that grades websites: eventually you have to point it at your own.
So we did. On July 8, 2026, we ran Sourceable's agent-readiness scanner against besourceable.com and published the result without editing it. The score was 78 out of 100 - Level 4, "Well-Mapped." Good, not great. And the way those points were distributed turned out to be more interesting than the number itself.
The scanner grades five pillars. Here's exactly what came back:
| Pillar | Score |
|---|---|
| Discoverability | 100 |
| Content Accessibility | 100 |
| Bot Access Control | 100 |
| Protocol Discovery | 10 |
| Commerce | not applicable |
| Overall |
78 / 100 (Level 4, "Well-Mapped") | Three perfect scores. One near-zero. That shape is the whole story, and we suspect it's the shape most sites would produce if they ran the same test today.
The first three pillars are what most people mean when they say "SEO basics," extended to AI crawlers.
Discoverability (100) means an AI agent can find our pages. Sitemaps, canonical structure, the machinery that says "here is what exists on this site."
Content Accessibility (100) means the agent can actually read what it finds. Our facts live in real, server-rendered text, not trapped inside images, PDFs, or JavaScript that only executes for a human browser. If a model wants to know what Sourceable does, it can extract that in plain text.
Bot Access Control (100) means we're deliberately letting the right crawlers in. Not blocking GPTBot by accident, not throttling ClaudeBot because of an old security rule someone set in 2023. This one is worth pausing on: a lot of sites score badly here not because they chose to block AI, but because a default setting or CDN rule quietly did it for them.
These three are table stakes, and they're achievable. If your site does nothing else, do these. They're the difference between an AI being able to cite you and an AI never seeing you at all.
Then there's the fourth pillar, where we scored a 10 out of 100.
Protocol Discovery measures something newer and stranger: whether your site exposes machine-readable descriptions of what an AI agent can do with you, not just read about you. Not "here is my content" but "here are my APIs, here is how to authenticate against them, here is what my agent can do, here are the skills you can call."
The scanner told us exactly what was missing to reach Level 5 ("Agent-Ready"): None of that existed on our site. We were, in the scanner's words, well-mapped for readers and nearly invisible to actors.
Because that 68-point gap between our worst pillar and our best three is not a Sourceable problem. It's the shape of the entire web right now.
The last two decades of web development optimized for one kind of visitor: a human with eyes, arriving through a browser. The last two years added a second: an AI that reads your text to answer someone's question. Both of those are passive consumers of content, and the first three pillars serve them well.
The thing almost nobody has built for is the third kind of visitor, the one arriving next: an agent that doesn't just read your site but does something with it. Compares. Books. Buys. Calls your API. That agent needs a machine-readable contract, not a beautifully written paragraph. It needs to discover your capabilities the way a crawler discovers your pages.
Right now that layer is nearly empty across the web. Which means it's also the cheapest competitive advantage available, because you are not fighting an incumbent for it. You are simply earlier.
If you run this scan on your own site, you'll likely see one of three patterns. Each one tells you something different about where to spend your next hour. Low on the first three pillars. Stop everything and fix this. If crawlers can't find your pages, can't read your facts, or are being blocked outright, nothing else you do in AI visibility matters. Your content never reaches the answer. This is a plumbing emergency, and it's usually fixable in a day.
Strong on the first three, near-zero on Protocol Discovery. This is us, and probably you. Your site is legible to readers, human and machine. It's illegible to actors. There's no fire here, but there is a window, and windows close.
Strong across the board. Rare. You're building for the agent web on purpose, and you're ahead of nearly everyone in your category.
We're closing our own gap in the order the scanner suggested, starting with the API Catalog and OAuth discovery, then the A2A Agent Card and Skills index. We'll publish the follow-up score when we get there, whatever it says.
That's the point of running the tool on yourself, honestly. A score you can't be embarrassed by isn't a diagnostic, it's a marketing asset. The 10 is the useful number. It's the one that told us what to build next.
The scan is free, it reads only public pages, and it takes seconds. You'll get a 0-100 score across the same five pillars, your level, and the specific list of fixes to reach the next one. Most teams discover at least one thing they didn't know was broken, and a surprising number discover they've been quietly blocking the crawlers whose answers they were hoping to appear in.
We scored 78. We'd genuinely like to know what you score.
Check whether your site is AI agent-ready with Sourceable What is agent-readiness?
It's how easily AI agents and assistants can find, read, and act on your website. Sourceable scores it 0-100 across five pillars: Discoverability, Content Accessibility, Bot Access Control, Protocol Discovery, and Commerce.
Is agent-readiness the same as SEO?
No. SEO optimizes your position on a results page. Agent-readiness measures whether a machine can access and use your site at all. The first three pillars overlap with technical SEO; Protocol Discovery has no SEO equivalent, because it describes what an agent can do with you, not just read.
What is Protocol Discovery?
The pillar covering machine-readable declarations of your capabilities: an API catalog, OAuth/OIDC discovery, protected-resource declarations, an A2A Agent Card, and an Agent Skills index. It's what lets an agent discover and call your services, rather than only reading your pages.
Do I need Protocol Discovery if I'm not an API business?
Less urgently, yes. If you have no public APIs or agent-callable services, the first three pillars carry almost all your AI visibility. But the agent web is arriving, and a machine-readable description of what you offer will matter sooner than most teams expect.
Why does Bot Access Control matter so much?
Because you can block AI crawlers without ever deciding to. Defaults change, CDNs add rules, and security teams tighten settings. If a model's crawler can't fetch your pages, your brand can't be cited in its answers. "We never blocked anything" is not the same as "nothing is blocked."