AI content and Google: penalties or not in 2026?

I am Louis Choquet, founder of Pharelia. In our work, the question of AI content and Google penalties comes up every week. This article addresses it using primary sources published in 2026: Google's official AI guide, the Quality Rater Guidelines, and core and spam updates. The goal is to explain what is penalised, what is not, and what to do.

No, not because it is AI. Google's position, voiced by Danny Sullivan, is consistent: assess the quality of the result, not how it was produced. There is no general penalty targeting generated text. What is penalised is low-value mass-produced content — whether a human or machine wrote it does not change the rating.

The Quality Rater Guidelines (September 2025 version, ~182 pages, ~16 000 evaluators) describe three spam categories: Expired Domain Abuse, Site Reputation Abuse and Scaled Content Abuse. A separate section targets low-effort, low-originality content. Text produced at scale without added value receives the lowest rating — this is where poorly used AI causes a website to fall.

The official AI guide, published on 15 May 2026, makes a clear statement: optimizing for generative AI means optimizing for search — therefore SEO. There is no separate discipline to invent. Its mythbusting section dismantles several beliefs circulating over the past year that lead to unnecessary work.

Search Engine Journal summarises the same idea: AEO and GEO remain SEO. The AEO / GEO / SEO divide is mainly marketing. Fundamentally, we optimize useful, sourced, clear content — for the nuance, see our note on AEO vs SEO vs GEO.

Yes, but towards quality, not a hunt for AI. The May 2026 core update ended on 2 June after ~12 days. The June 2026 spam update started on 24 June, globally and in all languages. The affected websites among our clients were publishing at volume without value — not using AI judiciously.

Google has repeated the same logic for years about AI-generated content: the production method is not the criterion. Reviewed AI content enriched with practical experience and verified data passes. Raw AI content published in bulk receives the lowest rating.

Google announced the launch of AI Overviews and AI Mode in France on July 22, 2026. AI Overviews appears when Google considers a generated answer useful for the query; it does not replace all conventional results.

The business impact on clicks is real. Ahrefs measures a CTR drop of 15 to 30 % on informational queries where AIO is deployed, and around 69 % zero-click queries. Another signal: AI Search traffic converts at ~14,2 % versus ~2,8 % for conventional organic traffic (Seer Interactive). Fewer clicks, but better-qualified clicks.

Search Console provides a generative AI performance report for AI Overviews and AI Mode impressions. Google reports a global rollout as of August 31, 2026, with access limitations depending on the property and impression volume. This data is also included in the Web report: do not add it to that report's total. In GA4, check the AI Assistant channel and source/medium data for identifiable sessions. Traffic with no identifiable source may appear as Direct — a custom channel group does not reconstruct a lost source. Do not therefore reattribute all Direct traffic to AI. For detailed context, see our guide to Google AI Overviews in France.

No, not quite anymore. Ranking well on Google helps, but AI engines build their own citation index, weakly correlated with the SERP. You therefore need to work on presence where LLMs look for sources, alongside organic search — this is the core of what is called GEO.

In practice, ChatGPT accounts for ~87 % of AI referral traffic (Conductor 2026), while Perplexity ranks 3rd (~7,7 %, Statcounter May 2026). Reddit is disproportionately cited by LLMs because its text can be fully retrieved — a distribution tactic to include. For foundational work, follow our SEO audit checklist.

By treating AI as an accelerator, not an autonomous generator. The survival rule fits in one sentence: publish nothing you could not sign with your name as an expert. Every page must contribute data, an example or a decision — otherwise it falls into the low-value category targeted by the spam update.

This is the method we apply at Pharelia: AI for speed, human expertise for value. If you want to delegate this planning, see our services or start with Let's talk about your project.