Google Just Named the SEO Liability Killing Blogs: Commodity Content

On April 21, 2026, Google's Danny Sullivan stood on stage at Search Central Live Toronto and used the phrase "commodity content" to describe what broad core updates keep punishing. His three-part test for what survives: unique, specific, authentic. Industry analysis of the March 2026 core update showed pages with proprietary data gained 15 to 25 percent visibility, while templated SaaS blog posts lost 30 to 50 percent and generic AI content farms lost up to 80 percent.

What Sullivan actually named on the Toronto stage

Sullivan's title quietly shifted earlier this year. He's now Google's Search Director, not just the Search Liaison, which is part of why the framing felt more deliberate this time. He didn't lean on "helpful content." He didn't lean on E-E-A-T. He gave marketers something specific enough to audit for on a Monday morning: commodity content.

From Barry Schwartz's recap of the Toronto event, the three criteria Sullivan put up were:

• Unique: brings a viewpoint, data, or information competitors can't easily replicate.
• Specific: talks about a particular instance or situation, not general rules or generic steps.
• Authentic: demonstrates first-hand knowledge or expertise.

The implication is less subtle than Google usually is. Content that scores 0 for 3 on this rubric is the content Google expects to be most easily replaced by AI Overviews, AI Mode, and direct answers. Sullivan's framing via Search Engine Journal was even blunter: "Your original voice is that thing that only you can provide. It's your particular take."

The grape juice test, and why it's sharper than E-E-A-T

The example Sullivan used is worth repeating because it's a better audit heuristic than most of what lives in the official guidelines.

The commodity version, for an interior designer: "2024 Kitchen Trends You Need to See," illustrated with green cabinet photos pulled from Pinterest.

The non-commodity version, same designer: "Marble vs. Grape Juice: Why I Refused to Install Stone for a Family of Five." Includes actual stain tests the designer ran with grape juice and turmeric, and a short video of why they rejected the client's request outright.

One post could be written by anyone with a ChatGPT account. The other couldn't exist without the designer. JC Chouinard's slide recap captured the full deck and Sullivan's other examples, which included padded "What time is the Super Bowl?" pages and sites built on word-game solutions that now get undercut when Google gives the answer directly.

E-E-A-T is a framework. Grape juice is a test you can apply in 15 seconds per URL. The audit question becomes: could this post exist if I (or my client) didn't exist? If yes, it belongs on the commodity shelf.

What to cut first from a typical SaaS blog

Here's the part a working practitioner can do before lunch.

From what I've seen in content audits after the March 2026 update, the easiest category to identify and cut is the template-driven listicle. Pages like:

• "10 Best [Tool] Alternatives" written without any actual usage notes.
• "[Keyword]: A Complete Guide" that quietly rewrites the top five SERP results.
• "How to [Do Basic Thing]" posts that exist to rank, not because anyone on the team has a real opinion on how to do it.

The data on this is ugly. ClickUp added 2,815 blog posts built on exactly this template and still lost 97.6% of its organic traffic, which is the clearest recent proof that volume of commodity content has become a liability rather than an asset. The pattern holds at the industry level too: small publishers running the same playbook lost 60% of their search traffic over the last 18 months, while large publishers with harder-to-replicate reporting lost 22%.

The NeuronWriter analysis of post-March-2026 winners and losers matches the Toronto message almost exactly. Proprietary data pages gained 15 to 25 percent visibility. Generic AI content farms lost up to 80 percent. The murky middle, meaning pages that rewrote other SERP results without adding anything, dropped 30 to 50 percent.

My blunt take: if you inherited a blog with 500 posts and 80% of them fit the template list above, the defensible move is probably to noindex the bottom half and redirect the writing budget into one real piece of first-hand content per week. Delete-to-rank sounds aggressive until you pull the crawl stats on those pages. Google is already rounding the crawl budget on most of them down to effectively zero.

Where AI writing still works (and where it stops)

I don't think this kills AI writing. It kills AI writing without inputs.

The part AI still does well: outlining, clean-up drafts, extracting structure from a messy transcript, turning bullet notes into readable prose. That's all useful, and it's not going anywhere.

The part that's now a visibility liability: asking a model to generate a post on a topic where the only inputs are training data plus the top 10 SERP results. On paper, that sounds like an upgrade for everyone. And sometimes it is. But it also means the practical cost of one ranking article just went up, because the prerequisite is now that somebody actually did the work (the test, the interview, the spend, the experiment) before the writing started. Teams that don't have a reliable supply chain for first-hand evidence are going to feel this most, and probably soon.

The Hobo analysis has been calling this "Helpful Content 2.0" for a few months, which I think is directionally right. The rubric has stayed roughly stable since 2022. What changed is that Google now has the generative infrastructure to actually enforce it at scale.

The SEO jobs that get harder after this

Commodity content was a real revenue model for a lot of agencies. If your pitch to clients was "we'll publish 40 SEO-optimized posts a month," Sullivan's Toronto slides took a blowtorch to the deliverable.

The labor is different now. It's less "operate the content machine" and more "be the researcher, the interviewer, the editor, and the person who actually ran the test." Which is mostly what good content strategists were doing already, minus the pressure to also hit a volume quota nobody really believed in.

The awkward truth is that the content teams who get hired next year probably won't be measured on posts shipped. They'll be measured on whether each post couldn't have been written by somebody else. That's a worse KPI to track in a spreadsheet and a better one for whether the page survives the next core update.

The audit to run before Friday

Open Search Console, filter pages by declining clicks over the last 90 days, and for each URL on that list ask one question: could a competitor with a ChatGPT account and a decent brief write the same page in 20 minutes? If yes, it goes on the kill list. If no, figure out what specifically makes it survivable (a quote, a chart, a proprietary stat, a lived test) and go make more pages like that.

It isn't a complicated audit. Most teams haven't run it because the old content playbook treated volume as a moat. Sullivan just said the quiet part out loud in Toronto: it isn't, and honestly, it hasn't been for a while.

Notice Me Senpai Editorial