What the 2026 core updates really changed for YMYL sites

We've had three core updates this year: a Discover update from 5 to 27 February, a broad update from 27 March to 8 April, and a second broad update from 21 May to 2 June (Search Engine Roundtable). Most people tracking them, Glenn Gabe included, rated March as fairly quiet and May as a big one.

If you work on health, trading or gambling sites, you probably felt May. And you're still living with it. Since June, Google has only shipped spam updates: a short one in June, another in August, and a long one from 24 September to 8 October (Dizains Web). Google doesn't announce core updates in advance, and recent gaps have run to six months, so May's results are what you're working with until the next one lands.

The short version of what I think happened: most of the YMYL sites that lost weren't untrustworthy. They were the wrong type of page for what Google now assumes the searcher wants. The rest of this post is the evidence for that, what it means in each vertical, and how we audit a drop.

What happened to YMYL

Gabe follows more sites through core updates than almost anyone. In May he saw health and finance start moving within a couple of days of the rollout. Gambling came a bit later and swung harder, across both casino and sportsbook domains. He summed it up as "YMYL content is held to a higher standard algorithmically" (GSQi). Nobody in these verticals will argue with that.

Hard numbers for gambling are scarcer, because the big public analyses mostly leave it out, and the best of them cover March rather than May. After the March update, an Affiverse study of more than 600,000 pages found 71% of monitored affiliate sites down, with typical losses of 20 to 35% (SharkLink). iGB Affiliate reported iGaming affiliates losing 20 to 35% on their strongest commercial pages, and several losing more than half their visibility on key money pages (iGB Affiliate). In May, gambling volatility spiked again on 23 May, 30 May and in the final tremor on 2 June. It's also telling that when Google rewrote its spam policy text on 15 May, both of its worked examples involved casino content (SharkLink, as above).

Who lost, and why

Those numbers tell you gambling got hit hard. They don't tell you why. The clearest public data on that is in health, from Aleyda Solis's Sistrix breakdown of the May update, because it shows which sites lost.

Look at who's at the top. GoodRx is a drug-price checker and Ubie is a symptom checker. Neither is a dodgy site. WHO and NIH, which are about as authoritative as sources get, slipped as well. The one that gained was WebMD, the site that tries to answer the whole question rather than one slice of it.

Outside health the pattern repeated. In the UK, canonical reference brands gained 24% while reference aggregators and tools lost 29%, and transactional marketplaces gained 21% (Aleyda Solis). Her line for YMYL is that trust gets you in the door, but "preferred page type and query intent still decide the outcome."

My reading is that this is search intent matching, but judged by the format of the page rather than its topic. Google reassigned which kind of page it treats as the default answer for each intent. The drug-price searcher didn't change. What Google assumes they want did. That's also why adding author bios and reviewer boxes to the wrong kind of page won't do much on its own. You should still have them. They just don't fix a format problem.

A caveat before going further: that chart is five health domains in one market, and a domain-level visibility score blends thousands of queries. WHO and NIH may have slipped for reasons that have nothing to do with format, and Solis herself flags parts of her data as possibly end-of-rollout noise. I'd treat this as a strong pattern to test against your own queries, not a law. The tooling and audit sections further down are how we do that.

How it plays out by vertical

Here's how the pattern plays out across the four verticals we work in, with a note on how solid the evidence is for each:

Two notes on the table. For gambling, Gabe also flags AI-generated and AI-translated content at scale, which Google treats as scaled content abuse. It's a common shortcut on casino sites that launch a lot of languages quickly, and I'd check for it first. For trading, Solis didn't cover finance, so that column leans on Sistrix's own write-up of the finished update (PPC Land) plus my own read of it.

Checking intent at scale

Everything above comes down to one question per query: what intent does Google now assume, and what kind of page does it reward for it? You can answer that by hand for 20 keywords. You can't for the thousands a multi-market casino or broker site ranks for, and that's where a keyword tool earns its keep. Most of the big ones now label intent; I'll use Semrush as the example. Its Keyword Magic Tool and Keyword Overview label every keyword as informational, navigational, commercial or transactional, and let you filter a whole list by intent. Semrush assigns those labels from the SERP features on the page, words like "buy" or "how" in the query, and branding (Semrush). The same intent data runs through Position Tracking and its competitive reports (Practical Ecommerce), so you can split a ranking drop by intent instead of staring at one blended line.

I treat those labels as a first sort, not the answer. A label tells you a keyword is informational. It doesn't tell you whether Google now wants a complete guide, a tool or a video for it, and in May that was the part that mattered. So filter by intent first, then open the results page for your biggest losers in each group and look at what kind of page is winning. That's your brief for what to build next.

The audit we run after a YMYL drop

1. Export the lost queries from Search Console. Looking at pages hides the pattern; queries show it.
2. Tag each query by intent (Semrush does this in bulk), then note what kind of source ranks for it now: official, tool, marketplace, video or reference.
3. Ask Solis's closing question: is our page still the best default destination for this query, in this market?
4. If Google now wants a different format, build it. Rewriting the byline on the wrong format is wasted effort.
5. On multilingual sites, go through every market for machine-translated or unreviewed pages, and recheck hreflang and canonicals while you're in there.

The mistake I see most often after a YMYL drop is keeping the page as it is and adding "more E-E-A-T". More often, the answer is a different page.

If your rankings held and only clicks fell, that's a separate problem, and Noah covers it in the 2027 case for SEO budget. If you'd like a second pair of eyes on a drop, that's what our algorithm recovery work is for.