Presented by AudioEye
AI agents struggle to read inaccessible websites. For retailers, that can quietly cost holiday sales before a shopper ever visits the site.
Shoppers are increasingly handing product research to AI agents, asking them to compare reviews and track down deals before they ever land on a retailer’s site.
Marketing teams have spent much of the past year figuring out how to show up in those results. But many haven’t asked a basic question: can AI agents actually read and use our site?
A recent AudioEye study suggests the answer depends on accessibility. We ran 1,560 AI agent sessions on six websites, each tested with and without accessibility fixes. On the site with the most accessibility issues, agents completed 96 percent of tasks with the fixes in place and just 31 percent without them.
So what does that mean for marketing teams in the middle of the holiday rush? Inaccessible sites can lose holiday sales from two sides. AI agents are less likely to recommend a product they couldn’t fully understand, so these sites can lose traffic from AI search. At the same time, shoppers with disabilities who do reach these pages run into the same issues. Some will leave without buying, and some may not come back.
AI agents can only recommend what they can read
AI search has quickly become a channel marketing leaders are trying to influence. Nearly half of U.S. adults have used an AI chatbot, according to a Pew Research Center survey. During the 2025 holiday season, traffic from generative AI sources to U.S. retail sites grew 693 percent year over year, according to Adobe Analytics.
But to recommend a product, an agent first has to read the page.
Many AI agents, including those in AudioEye’s study, do that through the page’s accessibility tree, a behind-the-scenes map of each page. Screen readers (software that reads a page aloud for people who are blind or have low vision) have used it for decades.
When the accessibility tree is missing information, the agent has to guess or give up. Some of the most common gaps happen because teams are moving too fast, or don’t know what to watch for. Size charts get uploaded as images because that's faster than building a table, but without a text description, an agent can't read the sizes. Color options often show up as small swatches, and without a text label, an agent can't tell navy from black.
None of that slows down a sighted shopper, but an agent looking for a navy wool coat in a size 16 that's on sale may come up empty. In our study, every model failed a task that required reading numbers shown only inside an image with no text description, including the most advanced models we tested.
When an agent can’t find what it needs, it may recommend a competitor’s product it can read instead. The shopper was ready to buy but the retailer never got a chance to sell, because the agent couldn’t read its page.
Shoppers with disabilities hit the same issues, and retailers can lose more than one sale
AI agents aren't the only visitors who depend on the accessibility tree. About one in four U.S. adults has a disability, according to the CDC. In the U.S., people with disabilities hold about $1.3 trillion in disposable income, according to the Return on Disability Group's 2024 Global Economics of Disability Report. That's a lot of holiday shoppers and a lot of holiday spending.
The same gaps that trip up AI agents can stop these shoppers too. Imagine a screen reader user trying to pick a coat color when none of the swatches have labels, or a keyboard user tabbing through checkout who never reaches a payment field that only responds to a mouse click.
A shopper who hits one of these issues can leave without telling the retailer what went wrong. For marketing teams, that shows up as a bounce on a product page or an abandoned cart, with no note attached. When conversions dip, the usual suspects get called in first: the messaging, the offer, the creative, the competition. Accessibility rarely makes the lineup, even when it may be the culprit.
The impact can outlast the season. A shopper who couldn't check out in December may not try again in January, and friends and family may hear about the retailer that worked better. The reverse holds too: a retailer whose site works for these shoppers has a real chance to keep their business long after the holidays.
A few quick fixes can still make a difference this season
For a team deep in holiday planning, the instinct might be “not now.” But the holidays are when these gaps cost the most. Traffic peaks, more of it arrives through AI tools, and every unlabeled swatch or image-only size chart turns away agents and shoppers at the moment they’re most ready to buy. The good news: some of the most useful fixes don’t need a code release, and your team can start on them this week
Start with your best sellers: On many platforms, image and product descriptions live in the content management system, so marketing teams may be able to update them without a code release. Add text descriptions to product images, and put size charts and sales prices on the page as text.
Run a quick check of key pages: Ask an AI agent to return a summary of your product page. Then run the page through AudioEye’s free accessibility scanner. Neither replaces a full audit, but together they can show which issues are quick to fix.
Give AI traffic its own channel: Break out visits from ChatGPT, Perplexity and other AI tools, and watch how they convert. That gives your team a baseline for next year.
Put the bigger fixes on your roadmap: Buttons, form fields and page structure usually live in site templates, so fixing one template can fix every page built from it. That work gets more urgent as agents move from research to checkout: Meta’s Muse, launched in September, can already fill out forms and make purchases for people.
Every holiday season, retailers compete for shoppers' attention. This year more of that attention runs through AI agents. The fixes that help an agent read a product page also help shoppers with disabilities find what they need and check out. That adds up to more holiday customers on both sides.
Chad Sollis is CMO at AudioEye.
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact sales@venturebeat.com.
