Google Content Quality in 2026
Google Content Quality in 2026:
What Got Punished and Why
Wikipedia lost 435 visibility points. Healthline dropped 21%. Affiliate sites were the most-hit category at 71%. Here’s the anatomy of what actually drove those losses — and the four quality signals now governing everything Google surfaces.
- The December 2025 Core Update (volatility: 8.7/10) wiped Wikipedia, Healthline, and the majority of UK news publishers — not for spam, but for failing to add genuine information gain to an already-indexed topic.
- February 2026 brought the first-ever Discover-specific core update. Your search rankings may be untouched; your Discover traffic may have collapsed. They’re now separate systems.
- Four signals explain the majority of December’s redistribution: information gain, E-E-A-T experience, topical authority, and authorship transparency. A site can fail any one independently.
- AI content penalty isn’t about AI. Unedited AI output lost 85–95% of traffic. Heavily human-edited AI-assisted content showed mixed results. The variable is information gain, not production method.
- The fastest recovery pattern seen in post-December audits: pruning thin supporting pages first, not rewriting cornerstone articles.
01Two Updates, Two Surfaces — One Quality Standard
Keep these two events straight. They’re related but they’re not the same thing, and conflating them leads to wrong diagnoses.
The December 2025 Core Update ran for 18 days and hit traditional SERP rankings across every vertical. SEMrush sensors rated it 8.7 out of 10 for volatility — the most disruptive broad core update of 2025. This is the event responsible for Wikipedia’s collapse, Healthline’s 21% traffic drop, and the affiliate site carnage.
Then on February 5, 2026, Google did something it had never done before: it announced a core update targeting only Google Discover — the interest-based mobile content feed — leaving traditional search rankings completely untouched. Barry Schwartz noted on Search Engine Roundtable that he could not recall Google ever doing a Discover-specific update of this nature. Alev Digital confirmed the separation: your keyword rankings in regular Google Search remain untouched.
Why this separation matters: Discover is no longer a passive downstream benefit of good SEO. Google just told us it has its own algorithmic governance layer — which means Discover drops can now occur even when your search rankings are holding steady. Publishers who’ve been treating Discover traffic as a bonus are now operating under a requirement they may not realize changed.
Both updates share the same underlying quality logic. But they operate on different surfaces with different inputs. The December data is your guide to traditional search. The February Discover data adds a new dimension. Understanding which one hit you — and how — is step one of any honest diagnosis.
02What the December Data Actually Shows
The loser list is specific enough to be genuinely instructive. Let’s start with the numbers.
Now the counterintuitive part. Wikipedia losing 435 visibility points isn’t a story about Wikipedia breaking spam rules or publishing bad content. It’s a story about Google recalibrating what “trustworthy” means — and that definition moved away from “large and well-known” toward “deep and verifiably expert in this specific thing.”
The winners make the logic even clearer.
The Greatschools vs. niche.com result is particularly clarifying. Two platforms covering adjacent subjects — one fell 21%, the other gained 13% — in the same update cycle, in the same category. That’s not a coincidence. That’s Google telling you directly what depth-over-breadth looks like in practice.
03The Four Signals Driving Quality Evaluation Now
Strip away the update-cycle noise, and four signals explain the majority of December’s redistribution. The important thing about this list: each signal can fail independently. A site with strong topical authority and weak authorship transparency can still take a big hit. December showed all four operating simultaneously.
Does this page add something new to the indexed record on this topic — or just restate what’s already there? This is the mechanism behind affiliate sites’ 71% negative impact rate. Most affiliate content is a recombination of specs and aggregated sentiment that’s already indexed elsewhere.
The January 2025 Quality Rater Guidelines update made enforcement materially stricter. Raters now assess E-E-A-T based on what the content demonstrates, not what the author bio claims. Fake experience signals — AI personas, inflated credentials, claimed firsthand knowledge the content contradicts — are now a specific target.
A cluster-level signal measuring depth and consistency of coverage on a subject area — not the quality of individual pages. A site can have dozens of individually adequate articles and still score poorly if those articles don’t interconnect and don’t cover subtopics comprehensively.
Google added a new “Authors” section to Search Central documentation in February 2026. Person schema with sameAs attributes linking to verifiable credentials creates a cross-reference path Google’s systems can follow. This is not decorative markup — it’s an active verification signal.
| Signal | What Google evaluates | December 2025 evidence |
|---|---|---|
| Information gain | Does this page add to the indexed record, or restate it? | Unedited AI content: 85–95% traffic loss (ALM Corp, 847 sites) |
| E-E-A-T (Experience) | Does the content itself demonstrate firsthand knowledge? | Medical pages without physician authorship lost symptom/treatment rankings |
| Topical authority | Is this site the specialist source on this subject area? | niche.com +13% vs. Greatschools.org −21% in the same category |
| Authorship transparency | Can Google verify the claimed expertise independently? | Google added “Authors” to Search Central docs, February 2026 |
04Information Gain: Why Wikipedia Lost and What That Means for You
Information gain is a signal rooted in a 2022 Google patent and operationalized in recent updates. The core idea: how much genuinely new information does this document add to the indexed record on this topic? Pages that restate what competitors already cover score poorly — regardless of length, formatting quality, or publishing frequency.
This is the mechanism behind Wikipedia’s loss, and it’s counterintuitive enough that it’s worth sitting with. Wikipedia has extraordinary domain authority. Its editorial standards are real. And yet: Wikipedia articles, by design, summarize the existing documented knowledge on a topic. They’re encyclopedic aggregations of what’s already known. When Google recalibrates its valuation of information gain, encyclopedic summaries get redistributed — not penalized, just redistributed toward sources that add something original.
The affiliate site picture is less sympathetic. 71% negatively affected. The reason: affiliate review content is, by construction, mostly a recombination of manufacturer specifications and aggregated user sentiment that’s already indexed everywhere. It adds almost nothing to the indexed record. ALM Corp’s 847-site analysis made the gradient visible:
The critical distinction: AI is not the target. Content that adds nothing to the indexed record is the target. You can produce that kind of content with a keyboard and a copy-paste habit just as efficiently as with a generative model. The production method doesn’t matter. The output does.
What does information gain look like in practice? Original data. Surveys with results nobody else has published. A/B test findings. A specific clinical case study. An industry practitioner’s firsthand account of something that happened — not synthesized from five other sources, but from direct experience. The thing that cannot be reproduced because it happened to you, or to your client, or to your organization specifically.
05E-E-A-T: The Experience Letter Now Has Actual Teeth
Google added a second “E” for Experience to its quality framework in 2022. For a couple of years, most publishers treated it as a soft signal — add an author bio, mention some credentials, move on. The January 2025 Quality Rater Guidelines update ended that era.
The key shift: raters are now instructed to assess E-E-A-T based on what the content itself demonstrates, not what the author bio claims. They added a specific new section targeting “fake E-E-A-T” — manufactured expertise signals including inflated credentials, AI-generated author personas, and claimed firsthand experience that the content itself contradicts. That last one is worth emphasizing. If your content reads like a synthesis of other sources but your bio says you’re a practicing clinician, the contradiction is detectable. And now it’s a target.
The Cleveland Clinic and Mayo Clinic declines are the sharpest illustration of this. These are institutions with genuine, unimpeachable medical expertise. If they took losses, the signal isn’t about who you are in the world. It’s about how clearly your content surfaces that expertise at the page level — specifically enough that both Google’s systems and human raters can verify it without leaving the page.
The YMYL expansion you may have missed: The September 2025 Quality Rater Guidelines revision extended high-E-E-A-T requirements beyond health and finance into elections, civic institutions, and public trust topics. Publishers covering government, policy, or institutional subjects who haven’t adapted to that expansion are operating under a requirement they may not know has changed.
What demonstrable experience actually looks like
Not: “Experts recommend a Mediterranean diet for cardiovascular health.”
Yes: “A 2024 meta-analysis in the Journal of the American College of Cardiology (Smith et al.) examined 15 randomized controlled trials with 12,461 participants. Adherence to a Mediterranean diet was associated with a 23% reduction in cardiovascular events over 5.2 years. The effect was strongest in European populations and less pronounced in other groups.”
The difference is specificity: named journal, named authors, specific study design, exact sample size, precise numbers, acknowledged limitations. That’s what genuine expertise looks like in writing. Not because Google requires a specific format — because that’s how someone who actually knows the field writes about it.
06Topical Authority: Depth Beats Scale
This one deserves a slow read, because it runs directly counter to what most content strategies were built on for the last decade.
Topical authority is a cluster-level quality signal. It evaluates the site’s overall depth and consistency of coverage on a subject area — not the quality of any individual page. A site can have dozens of individually adequate articles on a topic and still score poorly if those articles don’t interconnect, don’t cover the topic’s subtopics comprehensively, and don’t demonstrate sustained publication over time.
The December data surfaced this pattern across every category. Greatschools.org fell 21% while niche.com gained 13% in the same review-site vertical. Two adjacent platforms, opposite outcomes. The Google Discover update’s analysis by Clarity Global identified the same dynamic: “Expertise is now evaluated topic by topic, rather than at the domain level. A site with deep, consistent coverage of a single subject can outperform a bigger brand that only touches that topic once.”
The February 2026 Discover update added one more nuance that matters: Google explicitly confirmed that multi-topic sites aren’t automatically penalized. Their announcement included a sentence that directly contradicts the assumption that only single-topic specialists can win: “Since many sites demonstrate deep knowledge across a wide range of subjects, our systems are designed to identify expertise on a topic-by-topic basis.” So a local news site with a strong gardening section can surface in gardening queries. A movie review site that published one gardening article cannot. The signal is depth per topic, not depth per domain.
The practical implication: your internal linking architecture matters more than you’ve been treating it. Every article covering a subtopic should link to your pillar content on the parent topic, and pillar content should link back to the most substantive supporting pieces. Orphaned pages — articles that receive no internal links — contribute nothing to your topical cluster signal. They’re just noise.
08What the Discover Update Means for Content Strategy
The February 2026 Discover update differs from December’s broad core update in scope but shares its quality logic. Google’s announcement listed three priorities: surfacing more locally relevant content from publishers in the user’s country; reducing sensational and clickbait content; and highlighting “in-depth, original, and timely content from websites with expertise in a given area.”
The data from publishers is blunt. Search Engine Roundtable comment threads show operators reporting 90–95% Discover traffic drops, with one publisher citing a single-day loss of 90,000 clicks. Three categories are taking the heaviest damage: sensational headline publishers, thin-content aggregators, and international sites targeting US Discover audiences.
That last one matters for any non-US publisher that had built meaningful US Discover traffic. Google is now explicitly surfacing content from publishers based in the same country as the user. US Discover feeds are prioritizing American publishers. Non-US publishers should expect that pipeline to shrink until the international expansion of the update completes.
Context on the baseline: Reuters Institute data cited by Digiday showed Google Discover traffic was already down 21% year-over-year before the February update landed. The dedicated Discover update hit on top of an already-deteriorating baseline for feed-dependent publishers. If your content strategy had meaningful Discover dependence, both those signals apply simultaneously.
The convergence point, though, is actually clarifying for content strategy: the investments that strengthen traditional search performance — topical depth, named expert authorship, original reporting — are now the same ones governing Discover. Publishers don’t need two separate strategies. One higher editorial baseline, applied consistently, serves both surfaces.
09Where Google’s Quality Bar Is Heading in 2026
Read December 2025 and February 2026 as consecutive steps in a single trajectory. Aleyda Solís, international SEO consultant at Orainti, framed it well in her post-December analysis: the update “looks less like a traditional SEO reshuffle and more like Google hardening its rankings to support AI-driven search experiences.” That framing is precise. Content that clearly resolves user intent — niche, utility-led, verifiably authored — is structurally easier for Google’s systems to trust, excerpt, and surface in AI Overviews. Scale-driven output increasingly registers as noise.
Two patterns are converging to accelerate this. First, AI Overviews now intercept a significant share of informational query attention that previously reached organic results directly. Being cited in an AI Overview requires clearing a higher bar than traditional ranking: the content must be structured, authoritative, and independently verifiable. Second, the YMYL expansion in September 2025 pulled a much broader swath of content — government, policy, elections, civic institutions — into the strictest quality tier. Publishers covering those areas who haven’t adapted are operating under requirements they may not know changed.
The sites gaining ground in this environment share a consistent profile: narrower topical scope, greater depth, named and verifiable authors, and original data or firsthand experience woven into substantive pieces. That profile is not going to stop winning.
10What to Do This Week — Four Specific Actions
Not a 50-point checklist. The four things that the December 2025 evidence most directly supports, with specific tools attached to each.
Export your full page inventory from Google Search Console. Filter for pages averaging fewer than 10 impressions per month over the past 90 days. Any page in that bucket covering a topic already addressed by a higher-performing page is a candidate for consolidation or removal. The fastest recoveries seen in post-December audits came not from rewriting cornerstone articles but from pruning or consolidating the thin supporting pages diluting domain-level quality scores. Screaming Frog flags thin pages by word count in a single crawl; pair that with Search Console performance data to prioritize which to upgrade first.
This is the specific structured data move Google’s new “Authors” Search Central documentation points toward. WordPress users: configure Person schema including sameAs fields through Yoast SEO or RankMath’s author settings — both support this without custom code. The sameAs attributes should link to the author’s LinkedIn, published work on credentialed platforms, and any verifiable credential pages. Goal: a public entity graph connection Google’s systems can follow to cross-reference the claimed expertise.
Topical authority is a cluster-level signal. Every article covering a subtopic should link to your pillar content on the parent topic. Pillar content should link back to the most substantive supporting pieces. Ahrefs Site Audit and Semrush’s internal linking report both surface orphaned pages — articles receiving no internal links, contributing nothing to your cluster signal. Fixing orphaned pages is one of the fastest structural improvements available and directly addresses the topical authority deficit that drove December’s niche-versus-generalist divergence.
First: verify every specific factual claim against its source. Second: add at least one data point, case detail, or perspective not present in the top five Google results for the target query — something with information gain. Third: publish under a named author whose expertise in the topic is demonstrable through their public record, not just asserted in a bio. The December data calibrates the stakes precisely: 85–95% traffic loss for unedited output, mixed results for substantially human-edited. The editorial floor determines which side of that split you land on.
The bottom line this analysis keeps returning to: The Reuters Institute figure — organic search traffic already down 33% globally in one year — is not a market trend to monitor. For sites running on volume-based content strategy, it’s the cost of inaction, already running. In 2026, the question isn’t whether your content ranks. It’s whether Google’s systems trust it enough to surface it at all, on any surface it controls.
11Verified Sources
Every number in this post has a source. Here they are with honest notes on what each one is and what it isn’t.
Honest gaps in this analysis: We don’t have the exact percentage weight Google assigns to any of the four signals. The ALM Corp sample of 847 sites is meaningful but not a random sample. Discover traffic reporting from publishers is self-reported, not audited. The Reuters Institute/Chartbeat traffic decline data covers a panel, not the full web. Where confidence is lower, we’ve said so.
No affiliate links. No sponsored content. If something here is wrong or outdated, tell us and we’ll correct it with a public note. This post is also linked from our main content evaluation tools, where you can run your own site against the signals described above.
