How Many AI Images Have Been Created? The Numbers

Last updated September 22, 2026

There is no verified, up-to-date worldwide AI image total in the sources reviewed. Everypixel estimated more than 15 billion text-to-image outputs for 2022–2023. Later disclosures from Adobe and Google demonstrate scale but use different product and asset definitions, so they cannot be summed into a global count.

The useful answer depends on what is being counted: generated files, published images, editing operations or unique pictures. Private and local generation also fall outside many platform disclosures. The timeline below preserves dated evidence without projecting an old daily rate into today’s total.

Try it on Morphed: AI Image Generator.

Key statistics

15B+2022–2023
Historical AI image estimate

Everypixel’s original August 2023 analysis estimated outputs across several generators for 2022–2023. Its methods include extrapolation.

34M/day2023 estimate
Historical estimated daily rate

The rate in Everypixel’s analysis is historical. It is not a measured September 2026 worldwide rate.

1.5B+Feb 2026
Flow images and videos

Google disclosed cumulative mixed-format output in February 2026. It did not give an image-only breakdown in this announcement.

Dated AI image and asset milestones

Each row is a separate disclosure. There is intentionally no total row because these figures have different scopes.

Date / periodReported figureWhat it countsSource
2022–202315B+Estimated text-to-image outputs across selected generatorsEverypixel
February 202518B+Cumulative Firefly assetsAdobe
February 20261.5B+Cumulative Flow images and videosGoogle

The linked original sources define these metrics. A publication date is not evidence that the underlying count was measured again that year.

How has AI image volume grown year by year?

Available disclosures form a timeline, not a consistent annual series. Keep each figure attached to its product, output definition and reporting date.

18B+Feb 2025
Firefly generated assets

Adobe’s February 2025 announcement provides this dated platform milestone. Assets are not necessarily unique published images.

How we compiled this data

Reviewed September 22, 2026. We checked the original estimate and company disclosures, recording what each count includes. We did not annualize launch rates, invent missing years, deduplicate platform outputs or estimate private generation.

Before you cite these numbers

  • September 22, 2026 correction: the previous 80M images/day headline and 30B cumulative total were not established as current global measurements by their cited sources. They have been replaced with dated evidence.
  • Everypixel’s page now has a 2024 title, but its underlying analysis describes 2022–2023 activity and an original August 2023 publication. The historical estimate is labeled accordingly.
  • Firefly assets and Flow images-and-videos are not interchangeable with an image-only count. Overlap, revisions, retries and private outputs prevent a simple sum.
  • The absence of a reliable current total does not imply that generation has stopped growing; it limits what can be quantified from these disclosures.

Frequently asked questions

How many AI images have been created?

The sources reviewed do not establish a current global census. Everypixel estimated more than 15 billion text-to-image outputs for 2022–2023; later platform totals use different definitions.

Are 80 million AI images generated every day?

We could not substantiate that as a current worldwide measurement and have withdrawn it. Everypixel’s historical estimate is 34 million per day.

Does Google Flow’s total count only images?

No. Google’s February 2026 announcement reports more than 1.5 billion images and videos together.

Why not calculate a 2026 total from the historical daily rate?

That would assume a constant rate, complete coverage and no double counting. None of those assumptions is established, so such a calculation would be a scenario rather than a measured statistic.

Sources

Figures on this page are compiled from the following publishers and reports. Where sources disagree, we present the range and note the methodology difference.