AI Image Generation Statistics (2026)

A data analysis of how many images AI tools now generate, who uses them, what the market is worth, and why trust, labeling, and copyright should decide which generator you pick.

AI image generators now produce images in the billions. Adobe says its Firefly models have generated more than 24 billion assets as of June 2025 (Adobe). Google counted more than 5 billion Nano Banana images in the Gemini app by October 2025 (Google), and ChatGPT users made over 700 million images in a single week (TechCrunch). Yet people tell real images from AI ones only 62% of the time (Microsoft AI for Good). Below is the cited data, and our scored picks live in the AI image generation category.

Key takeaways

• More than 24 billion assets have been generated with Adobe Firefly as of June 2025, up from over 2 billion generations in September 2023 (Adobe 2025, Adobe 2023)
• 700 million+ images in one week: over 130 million ChatGPT users made them after the upgraded generator launched in March 2025, per OpenAI’s Brad Lightcap (TechCrunch, The Decoder)
• 5 billion+ Nano Banana images were generated in the Gemini app by October 2025 (Google), and Nano Banana Pro hit 1 billion images in 53 days (9to5Google)
• 86% of creators actively use creative generative AI, and 52% use it to generate new assets like images and video (Adobe)
• 62% accuracy is all people managed when sorting real from AI images across about 287,000 evaluations (Microsoft AI for Good)
• 76% of Americans say it is extremely or very important to know whether pictures, videos and text were made by AI, but 53% are not confident they can tell (Pew Research Center)
• Only 26% of people are comfortable with news outlets using AI to create a realistic image when no photo exists (Reuters Institute)
• 69% of creators worry about their content being used to train AI without permission (Adobe)
• $336.3 million vs $8.7 billion: that is how far apart two “AI image generator market” estimates sit, because they define the market differently (Global Market Insights, MarketsandMarkets)

How many images do AI image generators create?

Official counts now run into the tens of billions. Adobe says Firefly has generated more than 24 billion assets since launch (Adobe, 2025). Google reported more than 5 billion Nano Banana images in the Gemini app by October 2025 (Google), and ChatGPT users made over 700 million in one week.

No one publishes a global total. Vendor milestones are measured differently, so read them as floors, not a market total.

1. Firefly passed 2 billion generations during its beta, Adobe said in September 2023 (Adobe).
2. Firefly had generated over 7 billion images by April 2024 (Adobe). Volume more than tripled in seven months.
3. By October 2024, Firefly passed 13 billion images, an increase of more than 6 billion in six months (Adobe). That is roughly a billion new images a month from a single vendor.
4. Firefly reached over 18 billion assets by February 2025 (Adobe). From this point Adobe counts “assets”, because Firefly also generates video.
5. The count hit more than 22 billion assets, “including images and videos”, by April 2025 (Adobe). Later Firefly totals are therefore not image-only.
6. Firefly crossed 24 billion assets by June 2025 (Adobe). It is the latest count we could retrieve.
7. ChatGPT users generated more than 700 million images in about a week, with over 130 million users trying the upgraded generator after its March 25, 2025 launch, per OpenAI’s Brad Lightcap (TechCrunch; The Decoder). It is the only single-week count in this dataset.
8. Gemini app users generated more than 5 billion images with Nano Banana by October 13, 2025 (Google). Google added the model to the Gemini app on August 26, 2025 (Google), so that volume arrived in about seven weeks.
9. Nano Banana Pro reached 1 billion images in 53 days, Google’s Josh Woodward said in January 2026 (9to5Google). Free users get three Pro images a day, Google AI Pro subscribers 100, and Ultra subscribers 1,000.
10. Bing Image Creator users have made “billions of images” since March 2023, Microsoft said in December 2024 (Microsoft Bing). Microsoft did not give an exact figure.
11. Over 10 billion pieces of content carry Google’s SynthID watermark as of May 2025 (Google). SynthID also covers text, audio, and video, so this is not an image count.

We could not find an official generation count from Midjourney, and Canva newsroom pages could not be retrieved for verification, so neither appears in the totals.

Source: vendor newsrooms and blogs linked above; ChatGPT and Nano Banana Pro figures were stated by company executives and reported by TechCrunch and 9to5Google.

How fast is AI image generation growing?

Image making is the one media-creation use of AI that is clearly growing. The Reuters Institute found weekly use of AI to make an image rose from 5% to 9% across six countries between 2024 and 2025 (Reuters Institute). Adobe’s creative freemium monthly users, which include Firefly, passed 100 million in 2026 (Adobe).

12. Weekly use of AI for making an image rose from 5% to 9%, while video (3%) and audio (2%) generation stayed flat (Reuters Institute, 2025). The authors link the rise to wider free access to image features, for example in Gemini.
13. Weekly use of any generative AI tool nearly doubled from 18% to 34% in the same six-country survey (Reuters Institute). Image making is still a minority habit.
14. Adobe’s creative freemium monthly active users crossed 100 million, growing over 70% year over year, in its third fiscal quarter of 2026 (Adobe Q3 FY2026). That group covers Firefly, Express, and the web and mobile versions of Premiere, Photoshop and Lightroom.
15. Firefly ending ARR across the Firefly app and credit packs grew 40% quarter over quarter in the same period (Adobe Q3 FY2026). Adobe also said AI credit consumption was accelerating quarter on quarter.
16. Adobe’s AI-first ending ARR now exceeds $650 million, growing more than 150% year over year (Adobe Q3 FY2026). That figure spans creativity, productivity, and customer experience products, not just images.
17. Traffic to Firefly grew over 30% quarter over quarter, paid subscriptions nearly doubled, and first-time subscribers grew 30%, Adobe reported in June 2025 (Adobe). Usage was turning into paid seats and new customers.

What it means: casual use is still small, yet a handful of platforms hold enormous volume. The suite your team already pays for may include a capable generator.

How many creators and designers use AI image tools?

Most working creators now do. Adobe’s survey of more than 16,000 creators in eight countries found 86% actively use creative generative AI, and 52% use it to generate new assets like images and video (Adobe, 2025). Product designers are slower: only 31% use AI for core design work such as asset generation (Figma, 2025).

18. 86% of creators actively use creative generative AI, per Adobe’s Creators’ Toolkit Report, run with The Harris Poll across the US, UK, France, Germany, South Korea, Japan, India, and Australia (Adobe, October 2025). For content creators, AI is now part of the default toolkit.
19. The top uses are editing, upscaling, and enhancement (55%), generating new assets like images and video (52%), and ideation and brainstorming (48%) (Adobe). Editing real work beats generating from scratch, which favors tools that sit inside an existing editor.
20. 60% of creators used more than one creative generative AI tool in the past three months (Adobe). That is also how tool sprawl starts.
21. 81% say AI helps them create content they otherwise couldn’t have made, and 76% say it accelerated the growth of their business or follower base (Adobe).
22. The top adoption barriers are high cost (38%), unreliable output quality (34%), and uncertainty about how the AI model was trained (28%) (Adobe). Training provenance is now a purchase criterion, not just a legal footnote.
23. Only 31% of designers use AI in core design work like asset generation, versus 59% of developers using it for core development, in Figma’s survey of 2,500 users (Figma, April 2025). For designers, image generation is still a side tool, not the main job.
24. Designers report a 69% satisfaction rate with AI tools and 54% say it improves the quality of their work, below developers’ 82% satisfaction (Figma).
25. 78% of Figma’s respondents say AI significantly enhances their efficiency, but only 32% say they can rely on its output (Figma). Speed is accepted; trust in the result is not.

For more on how design teams pick and adopt tools, see our design tools statistics.

How are marketers using AI images?

Marketers have folded AI into production. HubSpot’s 2026 State of Marketing reports 80% of marketers use AI for content creation and 75% use it for media production (HubSpot). The tension is audience trust: 98% of consumers say authentic images and videos are pivotal in establishing trust (Getty Images, 2024).

26. 80% of marketers use AI for content creation and 75% use it for media production (HubSpot, 2026).
27. 61% of marketers believe marketing is experiencing its biggest disruption in 20 years because of AI (HubSpot).
28. 98% of consumers agree that “authentic” images and videos are pivotal in establishing trust, per Getty Images’ VisualGPS research covering over 30,000 adults in 25 countries from 2022 to 2024 (Getty Images). Getty sells both stock and AI imagery, so it has a stake in this finding.
29. 87% of respondents consider it important for an image to be authentic (Getty Images).
30. Almost 90% of consumers globally want to know whether an image was created using AI (Getty Images). Disclosure is an expectation, not a nice-to-have.

What it means: AI images are already in the marketing pipeline. Use them where realism is not the promise, such as concepts, backgrounds, and variations, and disclose where it matters. Our AI in marketing statistics cover the wider adoption picture.

How big is the AI image generator market?

It depends on who draws the boundary. Estimates under the same “AI image generator” label range from $336.3 million in 2023 (Global Market Insights) to $8.7 billion in 2024 (MarketsandMarkets), whose definition also covers video generation. Treat any single number as a scoping choice.

31. MarketsandMarkets projects growth from $8.7 billion in 2024 to $60.8 billion in 2030, a 38.2% CAGR (MarketsandMarkets). Its report scope includes video generation, video synthesis, and video editing, which inflates the “image” total.
32. Global Market Insights sized the market at $336.3 million in 2023, reaching $1.4 billion by 2032 at a 17.5% CAGR (GMI, July 2024). It uses a narrower, image-focused definition and gives North America around 35% of 2023 revenue.
33. The Business Research Company puts the market at $0.43 billion in 2025, rising to $0.97 billion in 2030 at a 17.5% CAGR (The Business Research Company). Two of the three firms agree on 17.5% annual growth, and it also names North America the largest region.
34. Adobe’s AI-first ARR alone exceeds $650 million (Adobe Q3 FY2026). It is broader than image generation, but it suggests the narrow estimates leave out revenue from AI features bundled into larger suites.

Source: forecaster pages linked above, retrieved October 2026. A fourth widely quoted estimate, from Grand View Research, blocked retrieval and is not included.

What it means: this is either a niche category or a multi-billion-dollar one, depending on whether you count video and bundled features. For buyers, the signal is that image generation is being absorbed into suites people already pay for.

Do people trust AI-generated images?

Not much, and least of all in news. In a 2025 Reuters Institute survey across six countries, only 26% were comfortable with outlets using AI to create a realistic image when no real photo exists, versus 55% for spelling and grammar edits (Reuters Institute). Comfort drops as AI moves from back office to what audiences see.

35. Comfort with AI in news falls as the output becomes visible: 55% are comfortable with AI editing spelling and grammar, 53% with translation, 47% with charts and infographics, 35% with a generic illustration, 26% with a realistic image, and 19% with an artificial presenter (Reuters Institute, 2025). Generic illustration earns more acceptance than photorealism.
36. Only 12% are comfortable with news made entirely by AI, versus 62% for entirely human-made news (Reuters Institute). The survey covered about 2,000 people in each of Argentina, Denmark, France, Japan, the UK, and the US.
37. 32% think journalists always or often use AI to create a realistic image when no photo exists, up from 28% in 2024 (Reuters Institute). Suspicion is rising faster than comfort.
38. 58% of people across 48 markets are worried about what is real and fake online when it comes to news, rising to 73% in the US (Reuters Institute Digital News Report 2025). Concern is lowest in Europe at 54%.
39. 50% of Americans are more concerned than excited about the increased use of AI in daily life, up from 37% in 2021; only 10% are more excited than concerned (Pew Research Center, 2025). The survey reached 5,023 US adults in June 2025.
40. 53% of Americans say AI will worsen people’s ability to think creatively, versus 16% who say it will improve it (Pew).
41. 76% of consumers agree “It’s getting to the point where I can’t tell if an image is real” (Getty Images, 2024). Doubt now attaches to real images too.

Can people tell AI images from real photos?

Barely. Microsoft’s AI for Good Lab analyzed about 287,000 image evaluations from more than 12,500 people and found an overall success rate of only 62%, slightly above chance (Microsoft AI for Good, 2025). An earlier PNAS study found people sorted AI-synthesized faces from real ones with 48.2% accuracy (PNAS, 2022).

42. People correctly sorted real and AI images 62% of the time across about 287,000 evaluations in Microsoft’s “Real or Not” experiment (Microsoft AI for Good). The study analyzed players from August 2024.
43. For AI-generated images alone, the success rate was 63%: 121,735 of 193,779 were identified (Microsoft AI for Good). More than a third of AI images passed as real.
44. Two generation types fooled people more often than not: GAN-made human faces and inpaintings both scored below 50% (Microsoft AI for Good). The hardest single AI image was spotted only 12.6% of the time. Players did best on images of people and worst on nature and urban scenes.
45. Microsoft’s in-development AI detector succeeded more than 95% of the time on both real and AI images (Microsoft AI for Good).
46. A 2022 PNAS study found 48.2% accuracy at spotting AI-synthesized faces, below the 50% of a coin flip (PNAS). Training and feedback lifted accuracy only to 59.0%.
47. The same study rated AI-synthesized faces 7.7% more trustworthy than real faces (PNAS, 2022).
48. A 2023 study of 3,002 participants in the USA, Germany, and China found state-of-the-art fakes “almost indistinguishable” from real media, with most participants simply guessing (Frank et al.). The result held across audio, image, and text.

What it means: visual inspection is not a verification method. If provenance matters to your work, it has to travel with the file.

How are AI images labeled and detected?

Labeling infrastructure is scaling faster than audiences notice it. Google says over 10 billion pieces of content carry its SynthID watermark (Google, 2025), and the Content Authenticity Initiative counts over 5,000 members (CAI). Yet only 19% of people see AI labels on news daily (Reuters Institute).

49. Google has watermarked over 10 billion pieces of content with SynthID as of May 2025 (Google). Invisible watermarks help checkers more than viewers.
50. SynthID verification in the Gemini app has been used over 20 million times since its November launch, Google said in February 2026 (Google).
51. The Content Authenticity Initiative, founded by Adobe in 2019, has over 5,000 members across media, technology, and civil society (CAI). Its Content Credentials, built on the C2PA standard, work as a cross-vendor provenance label.
52. In an Adobe study, 76% of US consumers said it is important to know if online content is AI-generated (Adobe, April 2024). Adobe attaches Content Credentials to Firefly output automatically.
53. Only 19% see AI labels on news daily and 28% weekly, while 77% use news daily (Reuters Institute, 2025). Labels exist, but most readers rarely meet them.
54. 53% of Americans are not too or not at all confident they can detect if something is made by AI (Pew).
55. The EU AI Act’s transparency rules come into effect in August 2026, the European Commission says, with a voluntary Code of Practice on marking and labelling AI-generated content (European Commission). Teams publishing AI imagery to EU audiences should plan for disclosure.

What are the copyright and commercial-safety risks?

Ownership is the open question. The US Copyright Office concluded in January 2025 that generative AI outputs are protected only where a human author has determined sufficient expressive elements, and not through prompts alone (US Copyright Office). Meanwhile, 69% of creators worry about their work training AI without permission (Adobe).

56. The Copyright Office’s Part 2 report, published January 29, 2025, rules out “the mere provision of prompts” as grounds for copyright (US Copyright Office). Human arrangement or modification of an output can still qualify, and AI assistance does not bar protection for a larger human work.
57. The Office’s AI inquiry drew over 10,000 public comments by December 2023 (US Copyright Office). A pre-publication Part 3 on AI training followed on May 9, 2025.
58. 69% of creators are concerned about their content being used to train AI without permission (Adobe, 2025). With 28% citing training uncertainty as a barrier, provenance is a buying issue.
59. Adobe says Firefly is trained only on content it has permission to use, including licensed Adobe Stock and public domain content, never on customer content (Adobe, February 2025). It is a vendor claim, but a specific one.
60. Getty Images says its generative AI tool is trained exclusively on permissioned content and offers indemnification on every image (Getty Images, 2024).

What it means: for internal mockups, rights questions barely matter. For ads, packaging, or anything you want to own, the training data and indemnity terms matter as much as the output.

What to do with this data

• Pick on rights before aesthetics for commercial work. With 69% of creators worried about training data (Adobe) and copyright tied to human authorship (US Copyright Office), favor generators that document their training sources and state indemnity terms for anything customer-facing.
• Standardize on one primary generator plus, at most, one specialist. 60% of creators already juggle more than one AI tool (Adobe). Choose a quality-first tool such as Midjourney or a workflow tool such as Leonardo AI, and check what your suite already includes, whether that is Canva or OpenAI’s image models inside ChatGPT.
• Keep a human hand in every final asset. Under the Copyright Office’s guidance, human arrangement or modification of AI output can qualify for protection; prompts alone cannot (US Copyright Office).
• Disclose AI imagery where trust is the product. Almost 90% of consumers want to know when an image is AI-made (Getty Images), and EU transparency rules apply from August 2026. Prefer tools that attach Content Credentials or watermarks automatically.
• Do not verify images by eye. At 62% human accuracy (Microsoft AI for Good), use provenance checks such as Content Credentials or SynthID instead.
• Ignore single market-size headlines. Base-year estimates differ by more than 20 times depending on scope. Judge tools on output, rights, and fit.

Compare our scored picks in the AI image generation category and the wider design category, and read how tools8020 scores tools. For adjacent data, see our design tools statistics and AI in marketing statistics.

Frequently asked questions

How many images has AI generated?

No one publishes a global total, but vendor counts are in the billions. Adobe Firefly passed 24 billion assets by June 2025 (Adobe), Gemini users made more than 5 billion Nano Banana images by October 2025 (Google), and ChatGPT users made 700 million in one week (TechCrunch).

What percentage of creators use AI image generators?

Adobe’s 2025 survey of more than 16,000 creators found 86% actively use creative generative AI, and 52% use it to generate new assets such as images and video (Adobe). Product designers lag: Figma found only 31% use AI for core design work like asset generation (Figma).

Can people tell if an image is AI-generated?

Not reliably. Microsoft’s AI for Good Lab found a 62% success rate across about 287,000 evaluations, slightly above chance (Microsoft). A 2022 PNAS study found 48.2% accuracy on AI-synthesized faces, and people rated those faces 7.7% more trustworthy than real ones (PNAS).

How big is the AI image generator market?

Estimates vary with scope. Global Market Insights put it at $336.3 million in 2023, reaching $1.4 billion by 2032 (GMI). MarketsandMarkets, which includes video generation, projects $8.7 billion in 2024 rising to $60.8 billion by 2030 (MarketsandMarkets). The gap reflects scope, not disagreement about demand.

Are AI-generated images copyrighted?

In the US, only partly. The Copyright Office concluded in January 2025 that AI outputs are protected only where a human author determined sufficient expressive elements, and prompts alone do not qualify (US Copyright Office). Its inquiry drew over 10,000 public comments (US Copyright Office).

Do consumers want AI images to be labeled?

Yes. Almost 90% of consumers globally want to know whether an image was created using AI (Getty Images), and 76% of Americans say it is extremely or very important to tell AI from human content (Pew). Only 19% see AI labels on news daily (Reuters Institute).

How we compiled these statistics

We drew these figures from 32 pages published by 19 organizations: vendor newsrooms and investor materials, survey publishers, academic papers, market research firms, and government bodies. Every figure was retrieved from the publisher’s own page or PDF in October 2026; two executive-stated milestones (ChatGPT and Nano Banana Pro) are cited through the press that reported them. Figures we could not retrieve, including Canva’s AI usage counts and Grand View Research’s market estimate, were omitted, and where sources disagree we show the range.

Sources

• Adobe: Firefly revolutionizes creative ideation with new mobile app (2025)
• Adobe: Adobe revolutionizes AI-assisted creativity with Firefly (2025)
• Adobe: Firefly app with commercially safe video model (2025)
• Adobe: Adobe launches Firefly Video Model (2024)
• Adobe: Adobe introduces Firefly Image 3 foundation model (2024)
• Adobe: New Firefly models and web app (2023)
• Adobe: Inaugural Creators’ Toolkit Report (2025)
• Adobe: Q3 FY2026 earnings script and slides (2026)
• Google: Nano Banana is coming to Google Search, NotebookLM and Photos (2025)
• Google: Image editing in Gemini gets a major upgrade (2025)
• Google: Nano Banana 2 (2026)
• Google: SynthID Detector (2025)
• 9to5Google: Gemini app has generated 1 billion Nano Banana Pro images (2026)
• TechCrunch: ChatGPT users have generated over 700M images since last week, OpenAI says (2025)
• The Decoder: ChatGPT’s image generation explodes with 700M creations in first week (2025)
• Microsoft Bing: Bringing more magic to Bing Image Creator (2024)
• Figma: 2025 AI report (2025)
• HubSpot: 2026 State of Marketing (2026)
• Getty Images: Nearly 90% of consumers want transparency on AI images (2024)
• MarketsandMarkets: AI image generator market worth $60.8 billion by 2030
• Global Market Insights: AI image generator market (2024)
• The Business Research Company: AI image generator market size forecast report 2026-2030 (2026)
• Reuters Institute: Generative AI and News Report 2025 (2025)
• Reuters Institute: Digital News Report 2025 (2025)
• Pew Research Center: How Americans view AI and its impact on people and society (2025)
• Microsoft AI for Good Lab: How good are humans at detecting AI-generated images? (2025)
• PNAS: AI-synthesized faces are indistinguishable from real faces and more trustworthy (2022)
• Frank et al.: A representative study on human detection of artificially generated media across countries (2023)
• Content Authenticity Initiative: Our members
• European Commission: AI Act regulatory framework
• US Copyright Office: Copyright Office releases Part 2 of AI report (2025)
• US Copyright Office: Copyright and artificial intelligence