Rich Results, Poor Test

Google’s announcement a few weeks ago that they’re phasing out the Structured Data Testing Tool in favor of a new, totally insular successor called the Rich Results Test definitely didn’t add any smiles to my summer.

The key difference between the tools is that this new one will not validate any structured data entities unless and until those entities drive rich results in Google Search. In other words, Google is deprecating a standards-based tool in favor of a results-based one, a change which, despite their predictable efforts to frame it as an “upgrade” (🥳), carries the message that structured data shouldn’t matter to you unless they generate rich results for your website in Google Search — that is, unless your search traffic can benefit directly from them today (and with all due thanks to Google, of course). Ask not what you can do for the semantic web; ask what the semantic web can do for you.

To those of us who have spent the last decade proselytizing for structured data markup not purely because it can reward your site with rich snippets in Google Search, but because it’s virtuous to contribute to a larger network of meaning being constructed across the web, this is a bummer. Google’s old tool tacitly endorsed our web-betterment philosophy, and this new one has no idea what we’re talking about.

The old Structured Data Testing Tool (visit that link while you still can) was Google’s original structured data validator, long admired by semantic web geeks for its broad scope and its value-neutrality. You could manually input, or link to a webpage containing, any object of structured data, and the tool would read it and tell you whether it conformed to the governing library’s standards, without further comment. It wouldn’t fail if you submitted a niche entity type; its aperture was wide enough to accommodate everything in the Schema.org vocabulary, no matter how marginal or new. It also wouldn’t fail if you submitted an entity written to a pre-Schema.org standard like RDFa or GoodRelations; the Schema.org database was expansive enough to cover those precursors, and that’s where the tool’s standards derived from. And, most significantly for UpBuild, it wouldn’t freak out if you tried to use or connect entities in an unorthodox way; on the contrary, I came to depend on it to legitimize every weird structured data markup experiment I ever performed.

For example: one time, I worked for a law firm whose clients were invited to leave reviews and ratings for the individual attorneys on the firm’s website. Since Schema.org doesn’t allow for the Person entity to take on Review or AggregateRating as properties, I tried to invert the whole structure, making AggregateRating the main entity, and attaching the subordinate Person entity as the ItemReviewed. It may not have made a whole lot of intuitive sense to a human coder looking it over, but you know what? The Structured Data Testing Tool proved it was valid. And soon after that, the attorney pages started showing up in search results with review/rating rich snippets (which the law firm of course saw as the whole objective, but which I saw as proof of my wizardly concept). We got a rich snippet out of this experiment, but that’s not the point; the point is that I probably would not have deployed such an avant-garde structured data tag to the live web had the SDTT not told me it was valid first.

Admittedly, the tool was never perfect. It would balk when you tried to connect linked tags across multiple pages using @id, refusing to crawl any external pages linked from the main page’s entity, and instead throwing an error on those lines. But who knows what kind of undue processing taxation the tool might have been subject to if that had been made possible and people had started using it to test masses of interconnected pages on enormous enterprise or university sites? This was an easy flaw to live with.

But this Rich Results Test is a different kettle of fish — you might say a smaller kettle with fewer fish. If you submit an entity for validation that doesn’t fall within the much narrower scope of structured data already returning rich snippets in Google Search (recipes, reviews/ratings, events, videos), the Rich Results Test just acts like it isn’t there.

Since I have TL;DR Marketing to thank for bringing this story to my attention in the first place, I’m going to link to their brief on the subject and embed the image of the test they ran on their own site, complete with welcome pictorial commentary, so you can see the difference for yourself:

“Blog? Service? Organization? Never heard of ‘em.”

Google has been at the center of the search world for decades and their market share could hardly increase further, but still, it always stings a bit when they find a new way to demonstrate that their centrality has gone to their heads. It’s hard to see a change like this and not wonder whether they’ve simply stopped caring about the wider web at this point (if they ever did), concerning themselves only with the web that their search engine serves up. Whether that’s what they believe or not, this “upgrade” certainly sends the message that it’s what you should believe.

Of course I get that they’re a for-profit company and don’t owe the wider web anything, but at times like these, I find myself wishing that they wouldn’t release these tools with beneficent applications in the first place if they’re just going to hobble them or take them away. This is the web. It’s increasingly where we live. Are we not supposed to be motivated to take care of it, enrich it, give of ourselves to it, without first being convinced that we’re going to get something in return? We’ve been applying that selfish attitude to planet earth, where we all quite literally live, for centuries, and look at the desperate state we’ve left that place in.

Anyway, now is a great time to explore alternative validators, ones that aren’t insular first-party corporate products with a vested interest in controlling the output, and that aren’t bound for the chopping block. My favorite is the Structured Data Linter, a creation of Gregg Kellogg. It’s free, open-source, easy to use, and enduringly committed to the semantic web as a critical element of the ongoing human project to organize the world.

What say you all? Am I overreacting? Or are you going to miss the old Structured Data Testing Tool as much as I will?

Related Posts

Leave a Reply

Your email address will not be published.