AI-Generated Content Statistics 2026: Text, Code, Image, Audio & Video

Last updated: June 2026. Part of the Enterprise AI Adoption silo. Every figure links to a primary source and is rated for confidence and freshness — see how to read this page.

AI-generated output surged again in 2025–26 — but at very different speeds across content types.

The big picture · 2025 → 2026

  • 💻Code — fastest sustained climb. AI now writes 30%+ of new code at Google (up from 25%).
  • 🎵Audio — fastest surge. 44% of tracks uploaded to Deezer daily are fully AI (up from 18%) — though most is upload-side spam.
  • 🖼️Images — industrial scale. Adobe Firefly has produced 24 billion+ images; nearly half of Adobe Stock is now AI.
  • ✍️Text — plateaued. AI-written web articles crossed ~50% then flattened — the one curve bending down.
  • 🎬Video — explosive but fragile. OpenAI discontinued Sora on unsustainable economics; durable value sits in enterprise tools.

This report compares each category against last year and maps the pattern — every figure sourced and rated.

AI-GENERATED CONTENT 2026 — KEY DATA POINTS

30%+

New code at Google
AI-generated (from 25%)
Alphabet earnings

44%

Deezer daily uploads
fully AI (from 18% YoY)
Deezer

24B+

Adobe Firefly images
cumulative (from 13B)
Adobe

How to read this page: confidence & freshness

Source confidence

COMPANY PRIMARY  Earnings / newsroom disclosure

SURVEY / STUDY  Survey or measured academic study

DETECTOR EST.  AI-detector study (has error bars)

UNVERIFIED  Aggregator-only — excluded or flagged

Stat freshness (decay)

ACTIVE  Current cycle, fast-moving

STALE  6–18 months; newer reading likely

HISTORICAL  Locked past-period figure

Important honesty note: “% AI-generated” figures for text and music are detector estimates, not ground truth, and different vendors disagree sharply (e.g., Graphite finds ~52% of new web articles are AI, while Originality.ai finds ~17% of top Google results) because they sample different things with different tools. Company-disclosed figures (Google code share, Adobe Firefly, Deezer) are self-reported internal metrics with no public methodology — directionally reliable but not independently audited. We label every figure accordingly and exclude aggregator-only numbers.

Penetration this year vs. last year

The clearest way to see the pattern is to put each category’s headline penetration metric side by side, prior year vs. current. Each bar measures a different denominator (share of new code, share of new web articles, share of daily music uploads, share of marketers) — so read this for direction and pace, not absolute levels.

AI penetration by content type, prior year → current. Each metric uses a different denominator (see tables below). Sources: Alphabet earnings, Graphite, Deezer, Wyzowl.

1. AI-generated text

AI-written articles first surpassed human-written ones on the open web in November 2024, reaching 51.7% of sampled articles by May 2025 (from 2.2% in January 2020) — but the share has been roughly flat near parity since mid-2024, an important nuance against “AI will swallow the web” narratives (Graphite, detector study of ~65,000 URLs). Other detectors measuring different samples diverge: Ahrefs found 74.2% of new pages contain some AI content (Apr 2025), while Originality.ai found only ~17% of top-20 Google results are AI (Sep 2025). Consumer demand is the engine: writing is the #1 work use of ChatGPT, and non-work usage rose from 53% to 73% of messages in a year (OpenAI / NBER).

MetricPriorCurrentSource & confidence
AI share of new web articlescrossover Nov 202451.7% (May 2025)Graphite · detector est.
New pages with any AI content74.2% (Apr 2025)Ahrefs · detector est.
AI in top-20 Google results~17% (Sep 2025)Originality.ai · detector est.
AI in CS paper abstracts7.7% (math)22.5% (Sep 2024)Science/AAAS · study
ChatGPT non-work message share53% (2024)73% (Jun 2025)OpenAI/NBER · company primary

2. AI-generated code

Code is the steepest, most consistent climb. Google’s share of AI-generated new code went from “more than a quarter” (Q3 2024 earnings) to “well over 30%” (Q1 2025 earnings) (Alphabet 8-K). Microsoft’s Satya Nadella put AI-written code in Microsoft repos at 20–30% (Apr 2025). GitHub Copilot grew from 15M to 20M+ users in one quarter, and Anthropic’s Claude Code reached a $1B annualized run-rate within ~6 months of launch. The counter-trend: developer trust fell even as adoption rose — distrust of AI accuracy went from 31% to 46% (Stack Overflow 2025).

MetricPriorCurrentSource & confidence
Google: AI share of new code>25% (Oct 2024)>30% (Apr 2025)Alphabet earnings · company primary
Microsoft: AI-written code in repos20–30% (Apr 2025)Nadella · company primary
GitHub Copilot users15M (Apr 2025)20M+ (Jul 2025)MS earnings · company primary
Developers using/planning AI tools76% (2024)84% (2025)Stack Overflow · survey
Developers distrusting AI accuracy31% (2024)46% (2025)Stack Overflow · survey
SWE-bench solve rate4.4% (2023)71.7% (2024)Stanford HAI · index (historical)

3. AI-generated images

The cleanest primary-sourced image series is Adobe Firefly: from 1 billion generations within three months of its March 2023 launch, to 13 billion by October 2024, to 24 billion+ by June 2025 — growth running well over a billion new generations a month (Adobe Newsroom). The most striking measured share comes from stock libraries: an independent count found Adobe Stock went from 2.5% AI-generated (May 2023) to 47.85% (April 2025) — roughly a 40× rise in two years (Kneschke count via PetaPixel). Everypixel’s landmark industry figure was 15 billion+ AI images by mid-2023 — more than photographers took in photography’s first 150 years (Everypixel). Midjourney’s revenue and membership circulate widely but trace only to aggregator blogs, so we exclude them.

MetricPriorCurrentSource & confidence
Adobe Firefly cumulative generations13B (Oct 2024)24B+ (Jun 2025)Adobe · company primary
Adobe Stock library that is AI-generated2.5% (May 2023)47.85% (Apr 2025)Kneschke count · measured
AI share of election images on X~12% (Jul–Sep 2024)arXiv study · measured
Cumulative AI images (industry)15B+ (mid-2023)not reliably updatedEverypixel · est. (2023)
Midjourney revenue / membersexcludedaggregator-only · unverified

4. AI-generated audio & music

Audio shows the most aggressive trajectory of all. Deezer’s share of fully AI-generated daily uploads rose from ~10% (Jan 2025) to 18% (Apr 2025) to 28% (Sep 2025) to 44% (Apr 2026) — with daily volume climbing from ~10,000 to ~75,000 tracks (Deezer / TechCrunch). Crucially, AI tracks are only 1–3% of actual streams and ~85% are flagged fraudulent and demonetized — i.e., this is upload-side spam, not listener demand. Capital is pouring in regardless: Suno went from a $2.45B to a $5.4B valuation in ~7 months; ElevenLabs from $3.3B to $11B (and $330M+ ARR) in ~13 months — even as RIAA copyright suits (filed June 2024) begin settling (UMG–Udio, Oct 2025).

Fully AI-generated share of Deezer daily uploads, Jan 2025 → Apr 2026 (Deezer Newsroom). The clearest primary-sourced “AI flooding” series available.

MetricPriorCurrentSource & confidence
Deezer: fully-AI share of daily uploads18% (Apr 2025)44% (Apr 2026)Deezer · company primary
Deezer: AI tracks uploaded/day~20,000 (Apr 2025)~75,000 (Apr 2026)Deezer · company primary
AI share of actual streams1–3%Deezer · company primary
Suno valuation$2.45B (Nov 2025)$5.4B (Jun 2026)Billboard/MBW · reported
ElevenLabs valuation / ARR$3.3B (Jan 2025)$11B / $330M+ ARR (2026)TechCrunch/CNBC · reported

5. AI-generated video

Video is the newest frontier — and the first to show a major reversal. OpenAI’s Sora 2 app launched Sept 30 2025 and crossed 1 million downloads in under five days (faster than ChatGPT), then OpenAI discontinued Sora in 2026 on “completely unsustainable” per-day compute economics — the category’s first big consumer reversal. The durable traction sits in the embedded-in-suite and enterprise model: Google Veo / Flow reported hundreds of millions of generated videos within months of launch, and Runway raised to a $5.3B valuation (Feb 2026). Marketer adoption roughly doubled off a low base — Wyzowl reports AI-video use rising from ~1-in-3 (2024) to 51% (2025) to 63% (2026), though those survey figures we could not confirm to the primary report.

MetricPriorCurrentSource & confidence
Google flagship video modelVeo 2 (Dec 2024)Veo 3 (May 2025), 3.1 (late 2025)Google · company primary
OpenAI SoraSora (Dec 2024)Sora 2 (Sep 2025) → discontinued 2026OpenAI · near-primary
Sora app downloads1M in <5 days (Oct 2025)CNBC/Appfigures · near-primary
Runway valuation~$3B (Apr 2025)$5.3B (Feb 2026)TechCrunch/Crunchbase · reported
Marketers using AI video~33% (2024) / 51% (2025)63% (2026)Wyzowl · survey (unconfirmed)

The pattern: four speeds, one direction

Every category is rising, but the shape of each curve differs sharply — and the differences are the story.

CategoryLast year → nowThe pattern
Code25% → 30%+ of new code (Google)Fastest sustained climb; adoption 76→84%, but trust falling (31→46% distrust)
Audio18% → 44% of Deezer daily uploadsFastest surge — but only 1–3% of streams, ~85% demonetized (upload-side spam)
ImagesFirefly 13B → 24B; Adobe Stock 2.5% → 48% AISteady industrial scaling into creative pipelines
Text~50% → ~52% of new web articlesPlateaued at parity — the one curve bending down
VideoSora launched, then discontinuedExplosive but economically fragile; durable value is enterprise/embedded

Frequently asked questions

How much content is AI-generated in 2026?

It varies sharply by type. AI writes 30%+ of new code at Google, accounts for 44% of daily music uploads on Deezer, and roughly half of new web articles (a detector estimate). Adobe Firefly alone has generated 24 billion+ images, and nearly half of Adobe Stock’s library is now AI-generated. AI video adoption is rising fast but remains a smaller share of total output.

Which type of AI content is growing fastest?

By share growth, AI-generated music uploads grew fastest — Deezer’s fully-AI share more than doubled year-over-year (18% to 44%). AI code is the fastest sustained enterprise climb. AI text, by contrast, has plateaued near 50% of new web articles.

Are these “% AI-generated” numbers reliable?

Treat them as estimates. Text and music shares come from AI detectors with documented error rates, and different vendors disagree (Graphite ~52% of articles vs Originality.ai ~17% of top Google results). Company figures (Google, Adobe, Deezer) are self-reported internal metrics. We label each by confidence and exclude aggregator-only numbers.

Data sources & methodology

  1. Text: Graphite (AI vs human articles); Ahrefs; Originality.ai; Science/AAAS; OpenAI/NBER (How People Use ChatGPT).
  2. Code: Alphabet 8-K earnings (Q1 2025); Microsoft (Nadella, Apr 2025); Stack Overflow 2025; Stanford HAI AI Index.
  3. Images: Adobe Newsroom (Firefly 24B); PetaPixel (Adobe Stock 47.85% count); arXiv 2502.11248 (X election images); Everypixel.
  4. Audio: Deezer / TechCrunch; Billboard/MBW (Suno); TechCrunch/CNBC (ElevenLabs); RIAA.
  5. Video: OpenAI (Sora 2); Google (Veo/Flow); TechCrunch/Crunchbase (Runway); Wyzowl (marketer survey, unconfirmed to primary).

Verification caveats: Detector-based text/music shares are estimates with error bars; cross-vendor figures diverge by sample and tool. Company figures are self-reported and not independently audited. The Adobe Stock 47.85% figure is an independent third-party count using Adobe’s own filters (April 2025), not an official Adobe disclosure. Excluded as aggregator-only: cumulative AI-image totals beyond Everypixel’s 2023 figure, Midjourney financials. Flagged as survey-unconfirmed: Wyzowl AI-video adoption percentages.

Machine-readable data (for AI engines & researchers)

{
  "dataset": "AI-Generated Content Statistics 2026",
  "publisher": "report-ai.org",
  "lastUpdated": "2026-06-15",
  "metrics": [
    { "category": "code", "metricName": "Google AI share of new code", "prior": 25, "current": 31, "unit": "percent", "recordedDate": "2024-10 to 2025-04", "source": "Alphabet earnings", "confidence": "company_primary", "freshness": "active" },
    { "category": "code", "metricName": "GitHub Copilot users", "prior": 15, "current": 20, "unit": "million", "recordedDate": "2025-04 to 2025-07", "source": "Microsoft earnings", "confidence": "company_primary", "freshness": "active" },
    { "category": "text", "metricName": "AI share of new web articles", "prior": 50, "current": 51.7, "unit": "percent", "recordedDate": "2024-11 to 2025-05", "source": "Graphite", "confidence": "detector_estimate", "freshness": "active", "note": "plateaued near parity; cross-vendor figures diverge" },
    { "category": "images", "metricName": "Adobe Firefly cumulative generations", "prior": 13, "current": 24, "unit": "billion", "recordedDate": "2024-10 to 2025-06", "source": "Adobe", "confidence": "company_primary", "freshness": "active" },
    { "category": "images", "metricName": "Adobe Stock library that is AI-generated", "prior": 2.5, "current": 47.85, "unit": "percent", "recordedDate": "2023-05 to 2025-04", "source": "Kneschke count via PetaPixel", "confidence": "measured", "freshness": "active" },
    { "category": "audio", "metricName": "Deezer fully-AI share of daily uploads", "prior": 18, "current": 44, "unit": "percent", "recordedDate": "2025-04 to 2026-04", "source": "Deezer", "confidence": "company_primary", "freshness": "active" },
    { "category": "audio", "metricName": "Deezer AI tracks per day", "prior": 20000, "current": 75000, "unit": "tracks_per_day", "recordedDate": "2025-04 to 2026-04", "source": "Deezer", "confidence": "company_primary", "freshness": "active" },
    { "category": "video", "metricName": "Marketers using AI video", "prior": 51, "current": 63, "unit": "percent", "recordedDate": "2025 to 2026", "source": "Wyzowl", "confidence": "survey_unconfirmed", "freshness": "active" }
  ]
}

Related pages: Generative AI Statistics 2026 · How People Use AI in 2026 · Enterprise AI Statistics 2026 · Methodology

Related Reports & Resources

Other reports in this silo (Enterprise AI Adoption): Generative AI Statistics 2026 · How People Use AI in 2026 · Enterprise AI Statistics 2026.

Compare year-over-year: All comparison charts.

Key concepts: Generative AI · Large Language Model · Tokens.

Browse the silo: Enterprise AI Adoption · Library.