Most Americans would let AI buy routine goods without approval, Croud finds

Why it matters

AI commerce will depend on what happens between a model recommendation and checkout. Brands need machine-readable product data and human proof that survives reviews, creator scrutiny and price checks.

Luke Smith and Ben Knight's Croud released a survey on August 11th finding that 69% of Americans are open to letting AI make a purchase without seeking approval in at least one category. The figure comes with a constraint that matters for anyone building agentic commerce: shoppers may delegate the transaction, yet many still want human evidence before accepting an AI recommendation.

The Croud Consumer Index is based on a survey of more than 2,000 U.S. consumers that Croud describes as nationally representative. According to the agency's announcement, three in four respondents would use AI-powered instant checkout in at least one product category, while half would delegate purchases across three or more categories.

Those numbers measure stated willingness rather than observed transactions. They still point to a meaningful change in how consumers imagine using AI. The shopper is moving from asking a chatbot what to buy toward authorizing software to complete the order, especially for routine purchases such as groceries and household essentials.

Smith, an ex-Google executive, and Knight, a former head of search, founded Croud in the U.K. in 2011 as a search performance agency. Their latest thesis extends that original business into a market where a model can assemble the product shortlist before a shopper visits a search engine, retailer or brand website. Croud now sells media, creative, analytics and AI-search services alongside proprietary software, giving the agency a clear commercial stake in convincing brands to rebuild their discovery strategies around AI.

AI can make the shortlist without closing the sale

Croud says 73% of AI users in its U.S. sample consult large language models before deciding on a specific brand or product. That puts ChatGPT, Gemini, Claude and similar interfaces upstream of the conventional search funnel, where they can define the consideration set before a shopper has formed a preference.

The recommendation does not end the research process. Among AI users, 39% said they checked an AI recommendation against at least four other sources before buying. Croud's report page says 43% turn to YouTube for validation, compared with 35% using Instagram or TikTok and 27% using Reddit.

That behavior creates a two-stage market. Models narrow the field using the product information, brand associations and third-party material available to them. Consumers then inspect reviews, creator videos, retailer pages and social discussions to decide whether the model's answer deserves trust.

Croud U.S. CEO Valerie Davis calls the space between recommendation and checkout the "validation economy." Davis joined Croud in May after leading Assembly Global's North American operation. Her appointment placed an executive with agency and retail experience, including work on Bloomingdale's e-commerce launch, over Croud's U.S. expansion as the agency pushes deeper into AI-mediated marketing.

For operators, the survey argues against treating AI visibility as another isolated acquisition channel. A product needs consistent descriptions, specifications, inventory information and use cases that a model can parse. The claims must also survive scrutiny when a buyer checks a retailer listing, watches a review or searches for complaints. Conflicting prices, stale product pages and vague return policies can break that chain after the model has already delivered a qualified prospect.

The checkout number carries a narrower message

The headline finding does not mean consumers are ready to give an agent unrestricted access to their wallets. Croud says most AI commerce captured by its research involves purchases below $100, with groceries and household goods leading the categories where respondents were comfortable with automated checkout.

That makes the initial opportunity look closer to delegated replenishment than autonomous luxury shopping. Agents can reorder products with familiar prices, predictable quality and manageable return risk before they take responsibility for expensive or subjective purchases.

Fashion complicates that pattern. Croud says AI-assisted fashion shoppers in its sample spent 56% more than non-users and were 23% more likely to search by style or aesthetic. The spending gap is an association within the survey, rather than evidence that AI caused shoppers to spend more. It still gives retailers a reason to support conversational queries built around an occasion, fit or visual identity instead of relying entirely on brand names and conventional keywords.

Autonomous checkout also raises product questions that the survey's willingness figures cannot settle. A usable purchasing agent needs current prices, inventory, delivery windows, return rules and permission boundaries. Consumers need controls for category limits, spending caps and exceptions. Merchants need reliable attribution when an AI interface recommends an item, validates it through outside sources and completes the transaction without a conventional visit to the storefront.

Croud is selling the response to its own findings

The report also functions as positioning for Croud's AI-search and measurement business. The agency points to its work with supplement brand Thorne, where Croud used its BrandCI and SEOCI tools to map citations and brand visibility across ChatGPT, Claude, Meta and Gemini.

Croud reports that Thorne recorded a 30% year-over-year increase in organic revenue, finished 22% ahead of its organic revenue target and reached a 66% LLM mention rate as measured by BrandCI. Those are company-supplied results, and the mention rate comes from Croud's proprietary measurement system. The program also bundled new content, technical website work, structured data and closer coordination between internal teams, which means the revenue change cannot be assigned to LLM visibility alone.

The case study nevertheless shows what agencies expect brands to buy next: systems that monitor how models describe a company, identify gaps in the underlying content and connect AI citations with commercial performance. Search optimization centered on ranking pages. AI optimization expands the job to managing the facts, entities and external proof a model uses to construct an answer.

Croud has capital behind that move. ECI Partners acquired a majority stake in the agency in 2024, replacing minority investor LDC. ECI said the investment would support acquisitions, international growth and further spending on Croud's technology. Croud says it now combines more than 600 employees with a network of roughly 2,900 marketing specialists.

The survey gives that strategy a demand-side argument. Consumers appear willing to let AI perform a larger share of shopping, while continuing to verify what the machine tells them. Brands therefore have to satisfy two evaluators at once: the model assembling the shortlist and the person looking for evidence that its recommendation can be trusted.