AI Fashion Photography Statistics 2026
Review by May 10, 2027, or sooner if a major source is updated.
Shoppers use product images to decide, and mismatched images show up in return reasons. Fashion executives are already using generative AI for image creation, while labelling is a separate trust and legal question. This page lists the verified figures, each with its source and date. Original AIORA research on this topic is coming soon.
AI fashion photography here means commercial fashion images produced with generative models, usually from a product photo the brand already has. The working description of that workflow, and where a studio shoot still leads, is on that guide and on the AI fashion photography page. This page does not repeat the definition. It collects the figures a journalist, an analyst, or a content lead can quote, and it says what each figure is not.
Each line names the organisation, the date, and links the document re-opened for this version. Salsify, Rithum, and Coresight Research are vendor-linked research, not government statistics. Where two sources disagree, both numbers stay.
How important are product images when shoppers decide to buy?
Product images and videos sit with titles and descriptions, just behind price and availability, in Salsify's 2025 shopper survey. The image share is not an outlier on that chart. Pricing and availability rank higher. Reviews rank lower. Read the 77% next to those companions, not alone.
Salsify sells product-content software, so this is vendor-linked research. The figures below are the report's own question, not the looser wording in secondary write-ups. Fieldwork was one day, 10 October 2024, on SurveyMonkey, stratified by age and gender. Completed responses: 1,910 (953 in the United States, 957 in the United Kingdom). The press release is dated 23 January 2025. The link is the report PDF.
- 84% of shoppers found pricing and availability "extremely" or "very" important when deciding to complete a purchase. Salsify, 2025 Consumer Research Report (PDF), fieldwork 10 October 2024, n=1,910.
- 77% found product titles and descriptions "extremely" or "very" important on the same question. Salsify, 2025 Consumer Research Report (PDF), fieldwork 10 October 2024, n=1,910.
- 77% found product images and videos "extremely" or "very" important when deciding to complete a purchase, the same share as product titles and descriptions. Salsify, 2025 Consumer Research Report (PDF), fieldwork 10 October 2024, n=1,910.
- 71% found customer ratings, reviews, and user-generated content "extremely" or "very" important on the same question. Salsify, 2025 Consumer Research Report (PDF), fieldwork 10 October 2024, n=1,910.
- 54% found enhanced product content, such as comparison charts, feature tours, and brand content, "extremely" or "very" important on the same question. Salsify, 2025 Consumer Research Report (PDF), fieldwork 10 October 2024, n=1,910.
The chart question was: when shopping online, how important are the following product page elements for helping you decide to complete a purchase? The base is all completed responses. The 77% for images is a stated importance, not a measured change in orders when images are added or removed.
How often do returns come from images or product content that don't match the item?
Published answers range from 11% to 71% because the questions are not the same question. One asks shoppers if bad content, including images, caused any return in the last year. One asks what share of return reasons was a description or photo mismatch. One asks companies what share of their returns they attribute to images that did not match the product. Quote the question with the percentage.
- 71% of shoppers had, in the last year, returned an item they purchased online because of incorrect product content (for example, images that didn't match the product, or an out-of-date product description). This is content plus images, not images alone. Salsify, 2025 Consumer Research Report (PDF), same sample as the importance chart, fieldwork 10 October 2024, n=1,910.
- 68% of consumers had returned clothing or footwear. Rithum, 2025 Global Returns & Profit Impact Report (PDF). Rithum commissioned what it describes as an independent survey of more than 6,000 consumers in the United States, Canada, the United Kingdom, France, and Germany. Vendor-commissioned, not a government series.
- 61% of consumers said wrong size or fit was the top reason they returned clothing or shoes. Rithum, 2025 Global Returns & Profit Impact Report (PDF), same survey. The key findings also call poor fit a primary return cause. This line uses the clothing-and-shoes wording.
- 33% said a returned item did not match the description or photos. Rithum, 2025 Global Returns & Profit Impact Report (PDF), same survey. Description and photos are one answer, not a photos-only rate.
- 69% of 190 surveyed brands and retailers in the United States and Europe, with revenues from $50 million to more than $5 billion, identified product returns as a significant business challenge. Coresight Research findings announced on PR Newswire, 17 January 2024. Report dated 6 December 2023. The report page is gated, so the public figures are from the announcement. Coresight is the researcher. The announcement was issued with a product-visualization partner. This page does not link that partner's product site. Vendor-linked research.
- Dissatisfaction with products not aligning with online images contributed to 11% of all returns among those surveyed companies. Coresight Research, announcement 17 January 2024, survey of 190 brands and retailers. This is a company-reported share of returns, not a shopper incidence rate.
- 63% said improving product-imagery quality was the top measure they planned in the next year to mitigate returns arising from inconsistencies in product imagery. Coresight Research, announcement 17 January 2024, same survey of 190 brands and retailers.
- 68% planned to invest in CGI in the next 12 months. Coresight Research, announcement 17 January 2024, same survey. A plan to invest is not a measured return reduction.
Do not average 71, 33, and 11. Salsify's 71% is a share of shoppers who had at least one content-related return. Rithum's 33% is a share citing photo or description mismatch, in a survey where size and fit is the larger clothing reason. Coresight's 11% is a share of returns, reported by companies, attributed to images that did not match. A reader who needs one number for "images cause returns" does not have one. Cite the question, the sample, and the date.
The Rithum PDF also cites older industry totals from an earlier National Retail Federation release and a retail index. Those cited totals are not Rithum's own measurement. This page does not repeat them. The industry size is the October 2025 NRF and Happy Returns release in the next section.
How large is the returns problem in retail, especially online and in fashion?
Returns are a large share of retail sales, and a larger share of online sales, before any image-specific cause is isolated. The current industry totals are from the National Retail Federation and Happy Returns. They are not fashion-only, and they are not a measure of bad photography.
- Retailers estimate that 15.8% of their annual sales will be returned in 2025, totalling $849.9 billion. National Retail Federation and Happy Returns, press release 15 October 2025, 2025 Retail Returns Landscape. Surveys in summer 2025: 2,006 US consumers who had returned at least one online purchase in the past 12 months, and 358 ecommerce professionals at large US merchants with over $500 million in revenue.
- An estimated 19.3% of online sales will be returned in 2025. National Retail Federation and Happy Returns, press release 15 October 2025, same report. Online, not all retail, and not limited to apparel.
- In 2024, returns were 16.9% of sales, or $890 billion. National Retail Federation and Happy Returns, press release 15 October 2025. That pair belongs to this release. It is not a figure to attribute to a vendor that cited an older NRF note.
Use the NRF figures for the size of returns. Use Salsify, Rithum, and Coresight when the question is how much of that pile is tied to images, descriptions, or fit. The $849.9 billion total is not a measure of bad product photos. In the Rithum survey cited above, 68% had returned clothing or footwear, and wrong size or fit was the larger clothing reason (61%).
Do on-model images help more than packshots or flat lays?
For apparel, accessories, and cosmetics, Baymard's usability research says a cut-out is often not enough. Shoppers use a human model to judge fit, length, and scale. The public article does not give a conversion percentage, and this page does not add one.
- For apparel, accessories, and cosmetics, products designed to be worn need a human model so shoppers can judge fit, length, and scale. Cut-out images alone leave those attributes hard to estimate. Baymard Institute, "Provide Images of Accessory, Apparel, and Cosmetic Products on a Human Model", published 1 December 2020. Large-scale usability testing. Qualitative guidance. No conversion percentage is stated on the public article.
- 31% of consumers in the Rithum survey would be less likely to return a product if it included real-life customer photos. Rithum, 2025 Global Returns & Profit Impact Report (PDF), same survey of more than 6,000 consumers. That supports showing the product in use. It is not a substitute for an on-model image, and it is not a test of AI models against photographed models.
As of 10 November 2026, no public, controlled test on fashion product pages was found that compares AI on-model images with photographs of a real model wearing the same garment. This page does not invent a conversion lift. It also does not promise that AIORA will publish one. The open question is real. The number is not.
A packshot or flat lay is the usual source file, and the right frame when the page needs the product alone. A ghost mannequin image is that product-only shape with the form removed. When to use it, and when the page needs a person, is ghost mannequin versus AI model photography. The product overview is ghost mannequin AI.
Catalog volume, including a season's intake, is covered in AI content production for seasonal collections. How luxury teams separate a campaign shoot from e-commerce volume is in AI fashion photography for luxury brands.
How widely has generative AI been adopted for fashion image creation?
The fashion-specific adoption figure that could be re-opened is from The State of Fashion 2026, by McKinsey and The Business of Fashion. It is a share of executives already using generative AI across several jobs, one of which is image creation. It is not a share of catalog images that are AI-generated.
- More than 35% of fashion and luxury executives report already using generative AI in areas such as online customer service, image creation, copywriting, consumer search, or product discovery. McKinsey & Company and The Business of Fashion, The State of Fashion 2026, McKinsey article 17 November 2025. The same sentence appears in the Business of Fashion chapter, 9 January 2026. The article's preceding sentence is about global fashion and luxury players. Do not read nearby cross-industry investment shares in that BoF chapter as a fashion-only rate. This page does not quote them.
- Zalando reported that generative AI reduced its image-production costs by 90%. This is one retailer's disclosure, not an industry rate, and it is not AIORA's −90% cost per image. The Business of Fashion, State of Fashion 2026 chapter, 9 January 2026.
- Zalando's vice president of content solutions said generative AI cut image production time from six to eight weeks to three to four days. The Business of Fashion, State of Fashion 2026 chapter, 9 January 2026. The time cut and the cost cut are both Zalando's.
- In the fourth quarter of 2024, 70% of Zalando's editorial content was AI-generated. The Business of Fashion, State of Fashion 2026 chapter, 9 January 2026. Editorial content, not every product image on the site, and not a figure for other retailers.
Neither 90% figure is an industry rate. The cost framework, using only the published AIORA figures, is in how to reduce fashion content production costs.
Do shoppers trust AI-generated product images, and do they need a label?
A 2026 experiment found that an "AI-generated" label changed how people judged a utilitarian product, and that the total effect on purchase intention was mixed. The products were a power bank and a perfume, in China. It is evidence about labels. It is not a fashion apparel product-page test.
- In a utilitarian product context, an "AI-generated" disclosure label reduced perceived authenticity, aesthetic appeal, and social presence relative to no label. Those negative pathways were not detectable in a hedonic context. The total effect on purchase intention was mixed, because a positive direct effect ran against the negative indirect effects. Yang, Liu, He, and Chu, Frontiers in Computer Science, 21 August 2026. Beyond the label itself: how disclosed content shapes consumer responses to AI-generated product imagery in E-commerce. 3×2 experiment, 276 valid responses, China, power bank versus perfume. The paper says "esthetic appeal". Not a fashion-apparel product-page test, and not restated here as a round percentage.
- Transparency obligations under Article 50 of Regulation (EU) 2024/1689, the EU AI Act, apply from 2 August 2026. Consolidated text of Regulation (EU) 2024/1689: Article 113 says the regulation applies from 2 August 2026, and the earlier and later dates in that article do not move Chapter IV, where Article 50 sits. The act as first published is at the Official Journal ELI. The European Commission note, 31 July 2026 says that from 2 August 2026 new transparency rules require certain AI systems to tell users when content has been generated or altered, including labels on deepfakes and machine-readable marks. This is a legal start date, not a survey statistic, and it is not legal advice.
What Article 50 means for a retailer publishing AI fashion images, including who marks the file and who writes a visible disclosure, is in EU AI Act labeling for AI fashion images. That guide is not legal advice either. Counsel decides whether a given frame needs a label.
A widely repeated "nearly 90% of consumers want transparency" line was not used. The original report behind it could not be re-opened for this version, so the figure is omitted rather than quoted from a secondary page.
What scale and cost figures can be stated for AIORA?
These are the only AIORA scale and cost figures on this page. They come from work with LuisaViaRoma. Marco Ritratti, former COO of LuisaViaRoma, founded AIORA Studio. The three figures are for that catalog. They are not an industry rate, and they are not Zalando's 90%.
- 10,000+ SKUs for LuisaViaRoma. LuisaViaRoma case study. AIORA Studio.
- €1.5M a year saved on that work. LuisaViaRoma case study.
- −90% cost per image on that catalog. Not Zalando's 90%, and not a rate for other brands. LuisaViaRoma case study.
Use the ROI calculator for your own counts. No LuisaViaRoma conversion or return rate is stated here.
Sources
- Salsify, 2025 Consumer Research Report (PDF). Fieldwork 10 October 2024. Press release 23 January 2025.
- National Retail Federation and Happy Returns, 15 October 2025.
- Rithum, 2025 Global Returns & Profit Impact Report (PDF).
- Coresight Research findings, PR Newswire, 17 January 2024. Report page: Coresight Research, report dated 6 December 2023, gated.
- Baymard Institute, 1 December 2020.
- McKinsey and The Business of Fashion, 17 November 2025. BoF chapter: 9 January 2026.
- Yang, Liu, He, and Chu, Frontiers in Computer Science, 21 August 2026.
- Regulation (EU) 2024/1689, ELI. Consolidated text. European Commission, 31 July 2026.
Frequently Asked Questions
On the verified sources re-opened for this page: Salsify found that 77% of shoppers rated product images and videos extremely or very important to completing a purchase (fieldwork 10 October 2024, n=1,910). The National Retail Federation and Happy Returns estimated that 15.8% of annual sales, $849.9 billion, would be returned in 2025, and 19.3% of online sales (press release 15 October 2025). McKinsey and The Business of Fashion reported that more than 35% of fashion and luxury executives already use generative AI in areas such as image creation (McKinsey article 17 November 2025). Article 50 transparency duties in the EU AI Act apply from 2 August 2026. AIORA's own published figures for LuisaViaRoma are 10,000+ SKUs, €1.5M a year saved, and −90% cost per image.
There is no single rate. Salsify found that 71% of shoppers had, in the last year, returned an online purchase because of incorrect product content, for example images that did not match the product or an out-of-date description (same 2025 report, n=1,910). That question covers content and images together, not images alone. Rithum's 2025 survey of more than 6,000 consumers found that 33% said a returned item did not match the description or photos. Coresight Research, in findings announced 17 January 2024, said products not aligning with online images contributed to 11% of returns among the 190 companies it surveyed. The questions and samples differ, so the percentages are not interchangeable.
This page cannot say. As of 10 November 2026, no public, controlled test on fashion product pages was found that compares AI on-model images with photographs of a real model in the same garment. Baymard Institute's 1 December 2020 article says apparel, accessories, and cosmetics need a human model so shoppers can judge fit, length, and scale. That research is qualitative and publishes no conversion percentage. This page does not invent a lift and does not promise a test.
They measure different things. Salsify asked shoppers whether incorrect product content, including images, had caused a return in the last year (71%, consumers, US and UK). Rithum asked why consumers returned items and recorded description-or-photo mismatch as 33%, beside a larger size-and-fit share (five countries). Coresight asked brands and retailers what share of their returns came from products that did not match online images (11% of returns among surveyed companies). A shopper share, a reason share, and a company-reported share of returns are not the same statistic. Where two reputable sources disagree, this page shows both.
Transparency obligations under Article 50 of Regulation (EU) 2024/1689 apply from 2 August 2026. The regulation applies from that date, and Article 50 is not in the chapters given an earlier or later start in Article 113 of the consolidated text. The European Commission says that from 2 August 2026, new transparency rules require certain AI systems to tell users when content has been generated or altered, including labels on deepfakes and machine-readable marks. Whether a given fashion image needs a visible label is a question for counsel. This page is not legal advice. The fashion-specific reading is at /guides/eu-ai-act-labeling-ai-fashion-images.