AI Content & Google's Quality Bar After September 2025

Google does not penalise content for being AI-generated. It penalises content that is thin, generic, unoriginal, or mass-produced without human value — whoever, or whatever, wrote it. That distinction is the single most important thing to understand about content in 2026. The production tool is not the issue; the absence of original value, expertise and human accountability is. Get this right and AI is a powerful editorial assistant. Get it wrong and you risk demotion — or, for the worst offenders, complete removal from search.

This guide sets out exactly where Google's quality bar sits after the September 2025 Quality Rater Guidelines update and the 2026 core updates, and how to publish AI-assisted content that survives and ranks. It sits in Unity SEO's Content Strategy cluster alongside the E-E-A-T guide and What Is Topical Authority.

What Actually Changed: A Timeline

• January 2025 — Scaled content abuse formalised. Google's Quality Rater Guidelines instructed raters to assign the "Lowest" rating to pages where all or almost all of the main content is AI-generated with little effort, originality, or added value (Contently). Three new spam categories landed: scaled content abuse, expired-domain abuse, and site-reputation abuse.
• September 11, 2025 — SQRG update (182 pages). Added evaluation criteria for AI Overviews (parallel to featured snippets), expanded YMYL to a new "Government, Civics & Society" category, and refined AI-content evaluation and reputation research (DEV Community / SQRG).
• March 2026 Core Update. Heavily rewarded Information Gain and first-hand authority; demoted aggregators, thin affiliate content and unreviewed AI content (NEURONwriter).
• May 15, 2026 — Spam policies extended to AI surfaces. Google clarified that all existing spam policies (cloaking, scaled content abuse, link spam, site-reputation abuse, doorway pages, thin affiliation) now explicitly apply to AI Overviews and AI Mode. Violations can trigger demotion or full removal (SEO Sherpa).
• May 2026 Core Update. Reinforced the same direction — first-party authority and genuine expertise over repackaged content (Innermedia).

The throughline is consistent: Google is building search around AI, and index quality is now directly tied to AI-answer quality. It rewards contribution over repetition.

The Rule in One Sentence

Per Google Search Central and the rater guidelines: appropriate use of AI is not against the guidelines, but using automation to generate many pages without adding value violates the scaled-content-abuse spam policy (ProAICraft). The question is never "Was AI used?" — it is "Does this page help users and deserve to exist?"

What Gets Penalised (the High-Risk Patterns)

The 2026 updates target identifiable patterns, not a tool (WebsCraft, Digital Dawn):

• Scaled content abuse — publishing many near-identical pages mainly to capture search traffic (e.g. the same service page across 30 towns with only the place name swapped). Since mid-2025, the worst cases face complete removal, not just a ranking drop.
• Low information density — many words, little unique information; "stretched" for volume.
• SERP-duplicate content — pages that merely rephrase the existing top results, adding no novel insight. Google now detects "rewrite of a rewrite."
• Template/structural sameness — identical H2s, identical "5-point lists," identical intros — recognised mass-production patterns.
• Missing author verification — anonymous content, or fake/inflated author personas, which warrant Low or Lowest ratings.
• Unreviewed AI on YMYL topics — health, finance, legal, safety, civic — judged most harshly; weak or generic copy here is dangerous.

Critically, Google now evaluates patterns across the whole site, not page by page — one weak section can drag down strong ones (WebsCraft).

What Survives and Wins

Google's own Nick Fox (SVP, Knowledge & Information) put it plainly at Google Marketing Live 2026: AI summaries raise the premium on deep content — material that goes "one or two levels deeper" than the AI's first-level framing (Lumar). Content that wins now:

• Adds new information to a topic (Information Gain) — original data, first-hand evidence, proprietary frameworks.
• Becomes a source worth citing, helping AI systems form their answers.
• Is difficult to compile from other pages — it can't have been written by just anyone.
• Carries an accountable, verifiable named author with genuine real-world expertise.

This is the same standard AI answer engines apply — they cite content that demonstrates expertise, not content that merely exists. Google and AI search are pushing in the same direction, which is why Unity SEO's Citation Triangle (statistics + cited sources + expert quotes) works for both.

A Practical Information-Gain Scorecard

Before publishing any page, score it (adapted from the March 2026 core-update analysis). Below 7 of 9 means rewrite or don't publish:

1. Proprietary data (0–2)
2. First-hand evidence / experience (0–2)
3. Original framework or point of view (0–2)
4. Expert byline with an external, verifiable profile (0–2)
5. Date stamp + refresh plan (0–1)

How to Publish AI-Assisted Content Safely

The dividing line between "AI-assisted" and "AI spam" is human editorial oversight (SiteOracle). Use AI as a tool, not a printing press:

1. Draft with AI, finish as a human. Add original analysis, examples and a genuine point of view.
2. Fact-check every claim against primary sources; never let AI fabricate quotes, statistics or sources.
3. Attach a real, named author with a linked bio and demonstrable expertise — required for E-E-A-T (see the E-E-A-T guide).
4. Never mass-produce. No publishing dozens of near-identical pages per day.
5. Apply extra scrutiny to YMYL (health, finance, legal, safety) — credentialed human review is non-negotiable.
6. Audit the whole site, not just new pages. Find and fix or remove any "Low"/"Lowest" pages — they drag site-wide quality down.
7. Keep unedited AI as a minority of output with human review on everything published.

Unity SEO produces AI-assisted content that meets Google's bar — expert-reviewed, original-data-led, and built to be cited. Part of the Content Strategy cluster.

Sources: SQRG (Sept 11 2025) overview; Contently on the Jan 2025 rater update; SEO Sherpa on May 15 2026 spam-policy extension; NEURONwriter on the March 2026 core update; Lumar / Google Marketing Live 2026; SiteOracle on site-reputation abuse.