A machine-readable website audit for executives

Open your website. Now imagine you cannot see it: no layout, no photography, no brand colors, only the code and text a program can parse. That blind visitor is now a large share of your traffic and an outsized share of your influence, because it includes the search crawlers, the AI platforms deciding what to say about you, and the agents acting for buyers. Your design budget is invisible to all of them.

Executives keep buying redesigns for the audience that sees and neglecting the audience that reads. The machine audience needs a different audit, and it is short enough to commission without a committee.

Check one: is the content there without JavaScript?

The first question is brutal and mechanical: when a machine requests a page, does the response contain the content, or a shell that assembles itself in the browser? Google's JavaScript SEO documentation describes the risk exactly: app-shell sites where “the initial HTML does not contain the actual content,” which forces a second rendering pass that classic crawlers get and many other bots never attempt. Google's own advice: “server-side or pre-rendering is still a great idea because it makes your website faster for users and crawlers, and not all bots can run JavaScript.”

The test is one command: fetch a key page with curl and search the response for your product name, your value proposition, your numbers. If they are missing from the raw HTML, the machine audience is reading an empty room, whatever the browser shows.

Check two: does every surface agree on who you are?

Entity consistency means every surface names and describes the same company the same way, and it is a facts problem before a markup problem. Your legal name, brand name, category, locations, and key claims should match across page copy, titles and descriptions, and Organization structured data. Google recommends the same name and identity properties everywhere, with sameAs links (pointers to your official profiles elsewhere) that confirm which company this is. Here is the failure in miniature, made up but typical: the homepage says “Acme Analytics”, the markup still says “Acme Software Group” from before the rebrand, and an AI system answering a buyer picks the markup's version and describes a company that no longer exists. The full method for finding these disagreements is its own discipline, covered in the website truth audit. The machine-readable audit consumes its output: one agreed set of facts, expressed identically to people and parsers.

Check three: does the markup tell the truth?

Structured data has policies, and they read like a claims audit. Google's structured data policies require markup to be “a true representation of the page content,” forbid marking up “content that is not visible to readers of the page,” and revoke rich-result eligibility for violations. And even perfect markup carries no promise: Google states it “does not guarantee that your structured data will show up in search results, even if your page is marked up correctly” according to its own testing tool.

So the audit checks three things: the markup validates, the markup matches the visible content, and the markup is current. Stale markup is the sneakiest failure, because it updates on template cycles rather than content cycles and quietly re-publishes last year's positioning to the most literal-minded readers you have.

Check four: are the discovery files coherent?

Robots directives (the rules telling crawlers what they may read), sitemaps, and canonical tags are the traffic signs of the machine layer. The audit checks them for coherence rather than existence. The sitemap lists the pages you actually want read and none of the retired ones. Robots rules block what you mean to block and nothing else. And canonical tags (each page's declaration of its one official URL) point where the redirects point. The migration checklist covers how these break in a rebuild; the audit checks them at rest.

The newer discovery files get a different label. Files like llms.txt and agent cards are conventions some sites publish for AI systems, and the honest status is experiment: Google says plainly that you do not need special AI files and that llms.txt neither helps nor harms your visibility in its search. Publish them if you like the bet, keep them consistent with everything else if you do, and never let anyone sell you one as a ranking device.

Check five: can a machine do anything here?

The last check looks forward: beyond reading, what can a machine do on your site? For most companies today the honest answer is “ask questions through a chat interface” or “nothing”, and nothing is an acceptable answer, chosen deliberately. If you do publish a callable interface, the audit asks the same questions as every other surface: is it discoverable, is it documented, does it answer with the same facts as the pages, and are its limits stated?

I publish one, so the example is inspectable: this site serves a machine-facing endpoint, described by an agent card, and the endpoint answers from the same approved content as the visible pages. Its limits are part of the publication: it is discoverable and callable, and formal protocol conformance is not claimed, because it has not been independently tested. That last sentence is the audit standard in miniature.

Reading the results

Score each check pass, fail, or experiment, and resist the urge to weight them into a number. The failures tell you the work: missing server-visible content routes to engineering, disagreeing facts route to the truth audit, lying markup routes to whoever owns templates, incoherent discovery files route to web operations, and the callable question routes to strategy. For the Google slice of the machine audience, the generative AI performance report in Search Console gives you a first-party readout of how the AI surfaces treat you, where the report is available. For the rest, measurement is its own discipline, and it starts with how to measure AI visibility.

Questions worth asking next

Sources

1. Google Search Central, structured data general policies. https://developers.google.com/search/docs/appearance/structured-data/sd-policies
2. Google Search Central, JavaScript SEO basics. https://developers.google.com/search/docs/crawling-indexing/javascript/javascript-seo-basics
3. Google Search Central, "AI features and your website" guidance. https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
4. Google Search Central, Organization structured data. https://developers.google.com/search/docs/appearance/structured-data/organization
5. Google Search Central Blog, "Introducing Search Generative AI performance reports in Search Console," June 2026. https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports