Can AI-written content earn authority? Yes — provided the human work shifts to where it now counts: framing, rigorous review, and expertise genuinely built into the text. Google confirmed this on May 15, 2026: its systems judge content quality, not how it was produced. An assisted, rigorous text earns authority; an assisted, lazy text shows on the page — and pays for it in visibility.
This note absorbs two earlier articles — "AI Is Reshuffling the Cards of Creation: 5 Trends for 2026" and "Top 5 AI Tools for Writers in 2026" — deindexed and redirected to this page.
The debate cooled between 2023 and 2026
In 2023, the fascination was with volume. Many companies published mass-generated content, barely reviewed, barely contextualized. These texts proved recognizable on the page, weak in search rankings, and fragile against successive Google updates that tightened the bar on genuine usefulness to the reader.
In 2024 and 2025, a more disciplined practice took hold in organizations that reinvested in review. The model shifted: AI produces structured raw material, and the writer takes that material and injects what can't be generated — the precise example, the brand's tone, cultural nuance, editorial stance, factual verification. Productivity rose without eliminating the writer; the role simply became more precise.
In 2026, Google's May 15 doctrine confirmed that the search engine's generative features rely on its core ranking and quality systems[1]. The direct consequence for writing: whatever penalized mediocre content in classic rankings now penalizes it in generated answers too. Selectivity is rising. Editorial laziness costs more.
What AI actually does well
Four contributions are tangible and stable. First, ideation: facing a new topic, a few minutes exchanging with a model is enough to map a thematic field, identify angles, and sketch a provisional outline — work that used to take hours of research. Then, reformulation: adapting a piece of content for several channels or audiences, notably into French, German and English — a common situation in Switzerland — takes minutes, with the writer left to fine-tune. The first structured draft: given a precise brief, the model delivers the skeleton of the final text — sections, transitions, aggregated facts. Finally, cross-checking: comparing sources, synthesizing divergent positions, flagging inconsistencies — under strict control, since models still invent references, and every factual claim must be verified at the source.
On the magnitude of the gain, there is solid data. A 2023 MIT study, in a controlled experimental setting using realistic professional writing tasks, measured an average 37% reduction in writing time with assistance, at equal or higher perceived quality according to independent evaluators[2]. That order of magnitude is useful for framing what's possible. It doesn't generalize: long-form content, demanding brand content, and strategic work fall outside the study's scope.
What models still don't do
The limitations observed since 2023 remain stable despite improving models. Narrative rooted in lived experience still escapes them: they produce correct stories, rarely memorable ones. Humor, irony and local references remain difficult — a writer from French-speaking Switzerland knows when to slip in a cantonal reference or calibrate a tone; models stay literal or borrow clumsily.
Swiss regulatory nuances — banking vocabulary, requirements under the FADP, wording compliant with the Code of Obligations — demand expertise models don't systematically have. And a firm stance, one that settles a question rather than cautiously balancing it, doesn't emerge without substantial framing: left to themselves, models drift toward consensus.
Where the writer's value shifts
A professional writer's value is no longer measured by words produced per hour. It's measured by the quality of the brief — audience, tone, stance, constraints, sources: a vague brief produces vague content. By the rigor of the review — one that merely removes errors lets generic tone and empty phrasing slide through; one that cuts, rewrites, and injects expertise transforms the text. By sector expertise brought in from outside — trust and fiduciary services, watchmaking, medtech, law: the precision a model can't correctly invent, and which Google recognizes within its E-E-A-T framework. And, finally, by editorial consistency held over time — a voice is built across dozens of consistent publications, not a single piece.
That's the chain I apply to the notes on this site: I frame the brief, a model clears the ground, I rewrite, I verify every claim at the source, I sign it. The time saved on raw material gets reinvested in what no model provides — the stance, the exact example, the Swiss context.
What the May 2026 doctrine changes for your authority
Two precise points. First, there is no specific mechanism aimed against assisted content: selection runs through the core ranking systems, and quality content is treated on its merits, assisted or not. Second, editorial authority signals now carry more weight: identified authorship, sources cited, content freshness. Author identification and the traceability of claims are made machine-readable through markup — I covered this in detail in Structured Data and AI Citability.
These self-contained, signed, verifiable pieces of content are the raw material of a website built to be cited; their actual uptake by models is then measured according to the protocol codified in MCVA Cahier No. 1. And it's the first thing I check in a citability-focused redesign: every page must be readable on its own, out of context, by a machine that decides in a few seconds whether to cite you.
Editorial discipline — a structured brief, an assisted draft, human enrichment, factual verification — is a matter of method, not technology. It separates organizations that maintain a reliable editorial presence from those that lose it. Writing, moreover, is just one of the fronts an SME has to manage in the face of AI; the bigger picture is laid out in What Should a Swiss SME Do About AI in 2026?.
Key takeaways
— The writer's value shifts from volume produced toward the brief, the review that transforms, and the expertise injected into the text. — Google judges content quality, not how it was produced: assisted and rigorous earns authority, assisted and lazy pays a price. — Editorial consistency held over time remains the one lever no model reproduces.
FAQ
Does Google penalize AI-written content? No, not as such. Since the May 15, 2026 doctrine, it's established that the core ranking systems judge content quality and usefulness regardless of how it was produced. It's mediocre content that gets filtered out — assisted or not.
How much time can AI-assisted writing actually save? The best available measure: a 2023 MIT study puts the average reduction at 37% of time on short professional writing tasks, at equal or higher quality. For long-form or strategic content, no honest generalization is possible yet.
Is a writer still necessary if AI writes correctly? Yes, but the role changes: framing, rigorous review, injecting sector expertise, and holding the voice's consistency over time. This work is what separates a correct text from one that earns authority — and it's exactly what selection systems reward.
How do you avoid the generic tone of AI-produced text? Through a precise brief, a review that cuts and rewrites rather than merely correcting, and by adding elements a model can't invent: real examples, sourced figures, a clear stance, local grounding. If a paragraph could appear on a competitor's site without anyone noticing, it needs rewriting.
Does your content already earn authority with AI systems? The Score GEO™ pre-audit measures, free of charge, where your company appears — and doesn't appear — in the answers from ChatGPT, Claude, Perplexity, Gemini and Mistral. A quantified answer within a few days. Request a free pre-audit
Sources
[1] Google Search Central, Optimizing your website for generative AI features on Google Search, published May 15, 2026. developers.google.com/search/docs/fundamentals/ai-optimization-guide [↩]
[2] MIT, Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence, 2023. economics.mit.edu/sites/default/files/inline-files/Noy_Zhang_1.pdf [↩]
Jérôme Deshaie is CEO and founder of MCVA Consulting SA, an augmented agency based in Valais, Switzerland. Fifteen years serving major international brands, now working directly with Swiss SMEs. Background.