Can an AI agent finish a purchase on your website? AudioEye, Inc. (Nasdaq: AEYE), the digital accessibility company, reported on September 24, 2026 that AI agents completed 31% of assigned tasks on the least accessible website in its test and 96% on an accessible version of the same site. NielsenIQ (NYSE: NIQ) reported the same day that 51% of US consumers used at least one AI-powered tool to support their shopping in the past month. Five other marketing technology and AI announcements carried September 24, 2026 dates. Adobe Inc. (Nasdaq: ADBE) brought Photoshop, Lightroom, Express and Firefly tools into Google Gemini and added Acrobat to its Claude plugin. Dataiku launched Agent Management, Anteriad released UK and EMEA B2B marketing survey results, EMARKETER acquired the research business of RetailX, and The Starr Conspiracy launched a free AI battle card tool.
NIQ’s number measures demand. AudioEye’s number measures whether a website can serve that demand once an agent arrives to act on it. In the marketing teams I work with, the first kind of number shows up in quarterly reviews and the second rarely exists.
NIQ’s Agentic Commerce Tracker, released September 24, 2026, put AI-assisted shopping at 51% of US consumers in the past month, with AI-powered product recommendations at 20% adoption and AI-powered personal shopping assistants at 16%. NIQ stated that the tracker draws on a monthly sample of approximately 500 US consumers. At that sample size a single month’s reading carries a margin of error of roughly plus or minus 4.4 percentage points, so the “first time above half” framing sits inside the noise. The trend line holds up better. NIQ’s May 5, 2026 release, using the same Quick Question method, reported 42%. A nine-point rise in about four months tells a marketing leader more than the 50% threshold does.
AudioEye’s September 24, 2026 study shows where that demand can stall. AudioEye, a digital accessibility platform (DAP) provider, sent 1,560 AI agents built on six commercial models to complete 13 real-world tasks on six websites, repeating each test ten times. On five retail, travel and direct-to-consumer sites, the only change between runs was whether AudioEye’s accessibility fixes loaded. AudioEye stated that agents read a page through its accessibility tree, the structured map of a page’s content that screen readers also use, and that task completion fell by about two-thirds for every model tested on its least accessible site, including the most advanced. In one task the figures an agent needed appeared only inside an image with no text description. Six models tried it ten times each, and all 60 attempts failed.
I’ve watched accessibility sit in the legal and compliance budget for a decade, funded to whatever the last demand letter required. AudioEye’s study moves the same backlog into agent readiness work. The alt text, form labels and button names a screen reader depends on are the same elements a shopping agent depends on, and your accessibility backlog now doubles as a conversion backlog. Nobody puts alt text in a keynote. It may still decide whether an agent can find your price.
AudioEye also reported that the median agent run used 43% more tokens when its fixes were absent, 128,000 against 90,000, and up to six times as many on the least accessible site. The agent’s operator pays for those LLM tokens. AudioEye’s study does not measure whether agent operators steer shoppers away from sites that cost more to read, and marketing teams can treat that as an open question to track through 2027.
Run the numbers on a mid-size retailer. Suppose the retailer expects 20,000 shopping agent sessions a month by Q1 2027, each trying to look up a product, confirm availability and reach checkout. At AudioEye’s 31% completion rate, 6,200 sessions finish the task. At 96%, 19,200 do. If 8% of completed sessions become orders at a $90 average order value, the retailer books 496 orders ($44,640) at the low rate and 1,536 orders ($138,240) at the high rate, a monthly gap of 1,040 orders and $93,600. AudioEye reported the 31% figure for its worst test site and did not publish completion rates for each of the other four commercial sites, so a given retailer’s gap may be much smaller. The only way to find out is to send an agent through the site and count.
Adobe’s September 24, 2026 announcement puts brand asset production inside the same assistants shoppers use. Adobe stated that tools from Photoshop, Lightroom, Adobe Express and Firefly are available inside Google Gemini across all Gemini plans, for enhancing photos, creating marketing assets and customizing designs. Adobe also added Acrobat to its Adobe for Claude plugin, bringing it to more than 80 creative and productivity tools, and added interactive editors for PDFs and Adobe Express designs. Guests can use the tools without signing in. A regional marketer or franchise owner can now restyle a set of product photos for a seasonal campaign inside a chat window, outside the digital asset management system and the approval workflow a brand team built. Your brand team has to decide which assistant-made assets can ship without review, and write that rule down before the holiday creative cycle starts.
Dataiku’s September 24, 2026 launch addresses the count underneath all of it. Dataiku stated that Agent Management inventories every AI agent an enterprise runs across AWS Bedrock, Databricks Agents, Google Vertex, Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce, Snowflake Cortex and Dataiku, measures business and technical performance, and flags the riskiest agents. Dataiku cited IBM’s AI in Motion research finding that fewer than one in five organizations keep a complete, current inventory of their AI systems. SiliconANGLE reported that Dataiku also published a Harris Poll survey of 685 chief information officers, conducted July 9 to 29, 2026, in which nine in 10 said they were confident they had complete tracking of their agents. The two surveys drew on different populations and definitions, so the gap between them is suggestive and unproven. An AI governance board needs a list of the agents marketing built in Agentforce or Copilot Studio before it can govern them.
Anteriad’s September 24, 2026 survey of 221 UK B2B marketers found that 81% reported revenue growth at their organization in the past year and 83% exceeded their marketing goals, including 24% who significantly exceeded them. Brahim Samoud, President EMEA at Anteriad, said more than half of UK marketers are not fully confident in their own budget process. AI channels without settled measurement add to that budget pressure, which is the case for putting agent completion rates and assistant-made asset volume into the budget model now.
EMARKETER and The Starr Conspiracy covered the research layer. EMARKETER acquired the research business of RetailX on September 24, 2026 and will publish The UK500 ranking of UK ecommerce retailers and The CustomerX Index, a pan-European cross-sector ranking launching in late September 2026, with new US editions to follow. The Starr Conspiracy launched Battle Card Builder the same day, a free tool that produces competitive intelligence on a company and three competitors in about four minutes. A four-minute battle card is a first draft. Someone in product marketing still has to check every claim against the competitor’s current pricing page before sales repeats it to a buyer.
Sort the seven September 24, 2026 announcements by who produced each number. NIQ measured consumers, and NIQ sells the agentic commerce measurement its tracker promotes. AudioEye measured agents on sites with and without its own product. Dataiku cited IBM’s research to size the problem its product addresses. Anteriad surveyed the B2B marketers it sells data and services to. Adobe, EMARKETER and The Starr Conspiracy published no performance figures.
Check who produced each number before you plan against it
NIQ’s 51% comes from a monthly sample of approximately 500 US consumers, and NIQ did not publish a margin of error, question wording or a definition of “AI-powered tool” in the release. AudioEye’s 31% and 96% completion rates come from its least accessible test site, AudioEye sells the fixes it tested, and the release reports the site-level results for that one site and the 68% maximum drop without per-site figures for the other five. AudioEye stated that an Ohio State University evaluator agreed with its scoring 95% of the time, which addresses scoring accuracy and leaves site selection with AudioEye. Dataiku’s inventory statistic comes from IBM research, and the CIO confidence figure comes from a Dataiku-commissioned Harris Poll survey reported by SiliconANGLE; neither measures Dataiku’s product. Anteriad’s revenue and goal figures are self-reported by 221 UK marketers and carry no stated margin of error. Adobe published no usage or customer figures for its Gemini or Claude integrations, and The Starr Conspiracy published no accuracy figures for its battle cards.
Every vendor in the set sells a product the number it published supports.
Strategic priorities for marketing leaders
- Run an agent task test on your top five journeys. Send an AI agent through product lookup, availability, pricing, account login and checkout ten times each, and record the completion rate for each journey.
- Move accessibility fixes into the conversion backlog. Rank missing image descriptions, unlabeled form fields and unnamed buttons by the revenue of the journey they block, and fund them from the growth budget.
- Read NIQ’s tracker as a trend line. Compare consecutive monthly readings and treat any single-month change of less than about four points as noise.
- Write approval rules for assets made inside assistants. Decide which Gemini- or Claude-generated images, PDFs and designs can ship without brand review, and require the rest to enter your asset management system before use.
- Inventory marketing’s agents before the governance board asks. List every agent your team built in Agentforce, Copilot Studio or any other platform, with an owner, a purpose and a performance measure for each.
- Verify AI-generated competitive intelligence before sales uses it. Assign a product marketer to check each battle card claim against the competitor’s current public materials.
Back to the opening question. Half of US consumers now use AI to help them shop, by NIQ’s count, and an agent that cannot read your product page cannot complete a purchase on it. A Chief Marketing Officer (CMO) who reports AI search visibility without an agent completion rate is reporting the front of the funnel and skipping the part that closes. This week, pick your five highest-revenue website journeys, send an AI agent through each one ten times, and put the completion rate next to conversion rate in your next operating review.
Here’s The News for September 24, 2026
AudioEye Study Finds up to 68% Drop in AI Agent Task Completion on Inaccessible Websites
AudioEye, Inc. (Nasdaq: AEYE), the digital accessibility company, published research on September 24, 2026 in which it sent 1,560 AI agents built on six commercial models to complete 13 real-world tasks on six websites, once with its accessibility fixes loaded and once without, repeating each test ten times. AudioEye reported that agents completed 96% of tasks on an accessible version of its least accessible test site and 31% on the inaccessible version, that task completion fell by about two-thirds for every model tested on that site, and that the median run used 43% more tokens without accessibility fixes (128,000 against 90,000). AudioEye reported that all 60 attempts at a task requiring numbers shown only inside an unlabeled image failed. Five test sites were retail, travel and direct-to-consumer commerce sites, the sixth was the W3C’s demonstration site, and AudioEye stated that WebJudge, an open-source evaluator from Ohio State University’s NLP group, agreed with its scoring 95% of the time.
Read the full press release (Published: September 24, 2026 | Source: PR Newswire)
Majority of U.S. Consumers Now Use AI to Shop, NIQ Finds
NielsenIQ (NYSE: NIQ), the consumer intelligence company, announced on September 24, 2026 findings from its Agentic Commerce Tracker showing that 51% of US consumers used at least one AI-powered tool to support shopping in the past month, the first reading above half. NIQ reported that AI-powered product recommendations are the most widely used application at 20% adoption, followed by AI-powered personal shopping assistants at 16%, and that adoption has grown steadily over the past two quarters. NIQ stated that the monthly tracker draws on its Quick Question research with a sample of approximately 500 US consumers, and that it will measure agentic commerce as a distinct channel within its retail measurement suite.
Read the full press release (Published: September 24, 2026 | Source: Business Wire)
Adobe Comes to Gemini and Expands What You Can Do in Claude
Adobe Inc. (Nasdaq: ADBE) announced on September 24, 2026 that tools from Photoshop, Lightroom, Adobe Express and Firefly are available inside Google Gemini across all Gemini plans, so users can enhance photos, create marketing assets and customize designs from a Gemini conversation. Adobe described the Gemini launch as the first step in its work with Google and said it plans to add more creative and productivity tools over time. Adobe also added Acrobat tools to its Adobe for Claude plugin for the first time, bringing the plugin to more than 80 creative and productivity tools across Claude and Claude Code on desktop, web and mobile, and introduced interactive editors for PDFs and layer-based Adobe Express designs. Users can start as guests or sign in with an Adobe account for higher usage limits and additional tools.
Read the full press release (Published: September 24, 2026 | Source: Adobe Blog (blog.adobe.com))
Dataiku Launches Agent Management to Track and Manage Performance of AI Agents Built and Running on Leading Platforms
Dataiku, the enterprise AI platform company, announced on September 24, 2026 at its Succeed conference in New York the launch of Agent Management, a standalone product that inventories the AI agents an enterprise runs regardless of the platform that built them, measures their business and technical performance, and flags the agents that carry the most risk. Dataiku stated that Agent Management connects to AWS Bedrock, Databricks Agents, Google Vertex, Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce, Snowflake Cortex and Dataiku, with OpenTelemetry support for custom environments. Dataiku cited IBM’s AI in Motion research finding that fewer than one in five organizations maintain a complete, current inventory of their AI systems. Agent Management will be generally available in October 2026, priced per instance annually with monitoring metered per agent.
Read the full press release (Published: September 24, 2026 | Source: Business Wire)
Anteriad Releases 2026 B2B Marketing Edge: UK/EMEA Report
Anteriad, the B2B marketing data and services company, released the 2026 B2B Marketing Edge: UK and EMEA report on September 24, 2026, based on its survey of 221 UK B2B marketers covering data, measurement, campaign strategy, budget justification and C-suite alignment. Anteriad reported that 81% of UK B2B marketers surveyed said their organization increased revenue in the past year and that 83% of marketing teams exceeded their goals, including 24% that significantly exceeded them. Brahim Samoud, President EMEA at Anteriad, said that more than half of UK marketers are not fully confident in their own budget process.
Read the full press release (Published: September 24, 2026 | Source: GlobeNewswire)
EMARKETER Acquires Research Arm of RetailX for Global Expansion of Trusted Retail Brand Rankings
EMARKETER, the digital marketing, media and commerce research company, announced on September 24, 2026 that it acquired the research business of RetailX, adding retail industry rankings to its retail and brand intelligence platform. EMARKETER stated that it will publish The UK500, the annual ranking of the United Kingdom’s top 500 ecommerce retailers, and The CustomerX Index, a pan-European cross-sector ranking set to launch in late September 2026, along with new editions covering the US market.
Read the full press release (Published: September 24, 2026 | Source: PR Newswire)
The Starr Conspiracy Launches Free AI Battle Card Builder for B2B Tech Companies
The Starr Conspiracy, a Fort Worth B2B marketing agency for technology companies, launched Battle Card Builder on September 24, 2026, a free tool that produces competitive intelligence on a company and three competitors in about four minutes. The Starr Conspiracy stated that the tool is built for B2B tech marketing and sales teams and runs on an autonomous AI research platform built by the agency’s own engineers.
Read the full press release (Published: September 24, 2026 | Source: PR Newswire)
Frequently asked questions
What marketing technology news was announced on September 24, 2026?
Seven marketing technology and AI announcements carried September 24, 2026 dates. AudioEye published a study of AI agent task completion on accessible and inaccessible websites, NielsenIQ (NIQ) reported that 51% of US consumers used an AI tool to shop in the past month, and Adobe brought its creative tools into Google Gemini and added Acrobat to its Claude plugin. Dataiku launched Agent Management, Anteriad released UK and EMEA B2B marketing survey results, EMARKETER acquired the research business of RetailX, and The Starr Conspiracy launched a free AI battle card tool.
What did AudioEye’s September 2026 AI agent study find?
AudioEye reported on September 24, 2026 that AI agents completed 31% of assigned tasks on the least accessible of its six test websites and 96% on an accessible version of the same site. AudioEye ran 1,560 agents built on six commercial models across 13 tasks, repeating each test ten times, and found that the median run used 43% more tokens when its accessibility fixes were absent. AudioEye sells the accessibility fixes it tested, and the release did not publish completion rates for each of the other test sites.
What share of US consumers use AI to shop, according to NIQ’s September 2026 data?
NielsenIQ’s Agentic Commerce Tracker, released September 24, 2026, found that 51% of US consumers used at least one AI-powered tool to support shopping in the past month. NIQ reported 20% adoption for AI-powered product recommendations and 16% for AI-powered personal shopping assistants. The tracker samples approximately 500 US consumers a month, which gives a single monthly reading a margin of error of roughly plus or minus 4.4 percentage points; NIQ reported 42% on the same measure in May 2026.
What did Adobe announce for Google Gemini and Claude on September 24, 2026?
Adobe announced on September 24, 2026 that tools from Photoshop, Lightroom, Adobe Express and Firefly are available inside Google Gemini across all Gemini plans for enhancing photos, creating marketing assets and customizing designs. Adobe also added Acrobat tools to its Adobe for Claude plugin, which now includes more than 80 creative and productivity tools, and introduced interactive editors for PDFs and Adobe Express designs inside Claude.
What is Dataiku Agent Management?
Dataiku Agent Management is a standalone product, announced September 24, 2026, that inventories the AI agents an enterprise runs across platforms including Salesforce Agentforce, Microsoft Copilot Studio, AWS Bedrock, Google Vertex, Databricks and Snowflake, measures their business and technical performance, and flags the highest-risk agents. Dataiku stated that it will be generally available in October 2026, priced per instance annually with monitoring metered per agent.
How can marketing leaders test whether AI agents can use their website?
Marketing leaders can send an AI agent through their five highest-revenue website journeys, such as product lookup, store availability and checkout, ten times each and record the share of tasks completed. AudioEye’s September 24, 2026 study found that missing text descriptions on images, unlabeled form fields and unnamed buttons stopped agents across every model it tested. Teams can then add the agent completion rate to the same report as conversion rate and prioritize accessibility fixes on the journeys where agents fail.
