About 19% of US adults, roughly 49.6 million people, now begin their retail product research with AI, according to PYMNTS research from August 2026. It changes what they buy. Among people who used AI to research products, 83% changed at least one purchase decision, and 27% chose a different brand.

The places those answers show up keep multiplying. In March, Shopify's Agentic Storefronts made millions of merchants' products discoverable in ChatGPT by default. Google followed in October, launching Universal Cart in the US across Search and the Gemini app, with checkout through the Universal Commerce Protocol. Personal agents like Meta's Muse now compare products and buy on people's behalf. On Meta’s third-quarter earnings call in October, Mark Zuckerberg said shopping is Muse’s most common use, with more than a quarter of its users shopping through it at an average purchase size of over $200.

Shoppers get fewer steps between a question and a checkout. For you, that means your product pages, feeds, and reviews do more of the selling before anyone visits your store. Profound’s data shows how your products appear in those answers, and its Agents handle much of the work to improve them, whether you build your own or start from our library.

Find the shopping prompts that matter for your catalog

Shoppers rarely ask AI to "buy running shoes." They ask which running shoes suit flat feet, how two models compare, or what to get a marathoner who has everything. In our Christmas in July study of 200,000 commercial prompts from the 2025 holiday week, 46.9% were product research and 22% were comparisons. Only 6.1% asked where to buy, and 7.4% came after the purchase.

So the prompts worth tracking describe needs, attributes, and comparisons, not only product names. Track them by category and by the attributes shoppers care about, like fit, material, compatibility, or price range, so you can see which parts of your catalog show up and which don't.

Profound gives you the data and an agent to build that set:

  • Prompt Volumes shows how real shoppers phrase questions about your category, with volume over time and intent.
  • Shopping tracks your products in ChatGPT Shopping. It shows how often your tracked prompts trigger a shopping answer, how visible each product and SKU is, how accurately the engine describes product attributes, shoppers' sentiment, and which merchant holds the checkout.
  • The Customer Question Miner agent reports the top questions asked about a topic, the underserved questions worth answering, and recommended content angles.

Create product pages AI can read and recommend

Your product pages carry most of what an engine needs to decide whether a product fits a shopper's question. The Christmas in July study compared the offers ChatGPT ranked first with the rest. Top offers were 188% more likely to have an FAQ section, 106% more likely to have product videos, 60% more likely to have customer Q&A, 35% more likely to have customer reviews, and 30% more images.

That’s a best practices playbook right there: write FAQ answers that help shoppers judge fit, sizing, compatibility, and use, and make sure reviews render in a format answer engines can read instead of loading through JavaScript. A fit guide for shoppers asking whether a product runs small is a good example, since it answers a question people ask before they'll commit.

Profound's agents help you make those changes across a catalog:

  • Research-backed PDP Optimizer uses our research team's latest findings, along with citation data and live search results, to recommend product page changes.
  • AEO-Optimized FAQ Generator builds an FAQ section from the query a page answers, Google's People Also Ask questions, and related sub-queries.
  • Optimize a Shopify Page for AI Search reads a page on your Shopify store, like a size guide or a shipping page, and drafts improvements for clarity and AI search. A person reviews and edits the draft, and only the changed fields are written back. If you reject the draft, nothing on your store changes.
  • Optimize a Shopify Blog Article for AI Search does the same for buying guides and other articles on your store's blog.

Pages also need to work for agents that browse on a shopper's behalf. Our research on personal agents found that agents like Muse navigate pages through the accessibility tree, the structure a browser builds from your HTML and labels, so semantic HTML and clear labels make your site easier for them to use. The same research recommends stating who each product is for, along with its intended uses, compatibility, dimensions, materials, and limitations.

Keep your feeds and product pages aligned

AI shopping engines draw on two kinds of product data: the feeds and catalogs you send them, and the product pages they crawl. In our June 2026 analysis of ChatGPT Shopping, the two supplied different details. Brand names and checkout images came from feeds, while delivery information and online availability came mostly from product pages. ChatGPT has leaned much harder on feeds since then, but the June data shows why both matter, since each supplied details the other didn't.

The feed side is getting easier to switch on. Shopify enrolls merchants in its agentic channels through Shopify Catalog, and brands that don't run their store on Shopify can join through its Agentic plan. Google's Universal Cart draws on its Shopping Graph of more than 60 billion product listings, with checkout through the Universal Commerce Protocol and the merchant as seller of record.

What gets harder is keeping every source consistent. A price updated on the product page but not in the feed, or a size that's out of stock in one and available in the other, gives the engine two versions of the truth. The Christmas in July study recommends auditing your feed for outdated prices and missing fields. Profound's Shopping view helps you spot the mismatches by showing how the engine describes your products' attributes and which merchant ends up with the checkout, whether that's your own store or a retailer that carries you.

Keep product facts accurate across AI answers

Wrong product details cost you twice: the shopper receives bad information, and they hold you responsible. In a Rithum survey of 1,046 US and UK shoppers, 58% blamed the retailer or brand when an AI recommendation had wrong product information, and 67% said price is the most important detail AI has to get right.

FactCheck finds those errors. It pulls the factual claims AI engines make about your brand and products, checks each one against your knowledge base, and traces every inaccurate claim to the page producing it, whether that's your site, a retailer's listing, or a review article. Your product facts live in Context Manager and your knowledge base, so every check and every agent works from the same approved specs.

The FactCheck Action agents take care of the solution, depending on where the error comes from:

  • Update Content revises your own page when it's the source of the wrong claim.
  • Create Content drafts new content when nothing you've published answers the question clearly.
  • Draft Outreach writes to a third-party site and asks it to correct the claim.

WHOOP built this into its workflows. The team uses FactCheck to monitor AI answers for inaccurate product claims and route them for correction, and FactCheck flagged 7.9% of AI responses for accuracy issues.

Earn the reviews, threads, and editorial picks AI pulls from

Shoppers use AI to understand other shoppers. In an IBM and NRF study of more than 18,000 consumers in 23 countries, 33% used AI to interpret reviews. Engines can also draw on gift guides, roundups, and community threads when they decide what to recommend, so the work extends well past your own pages.

Profound connects that work to what the engines cite:

  • Reviews. The Bazaarvoice integration connects Aim to Bazaarvoice's review tools. Aim monitors AI platforms for the topics and product attributes the engines need, and you can use those insights to trigger review requests, product sampling, and user-generated content campaigns through Bazaarvoice, with syndication across more than 11,500 retail and brand sites.
  • Editorial picks. The Citation-Friendly PR Target Finder agent pulls the domains AI cites in your category, checks whether each one allows AI crawlers, and ranks them into a pitch list.
  • Community threads. The FactCheck Action – Reddit Response agent drafts a reply to a thread that's spreading an inaccurate claim about your product, for your team to review before anyone posts it.
  • Sentiment. Shopping tracks shopper sentiment for your products, and the Top Sentiment Themes Report agent ranks the themes in what AI says about you, so you know which complaints to address first.

Measure AI shopping traffic and revenue

Agent Analytics shows when AI crawlers visit your pages, how many people arrive from answer engines, and, via its GA4 integration, brings in conversions and revenue. On Shopify, it connects through Nostra, or through a log drain or trace exporter on Plus and Hydrogen stores. Shopify also shows ChatGPT referral attribution on orders that come through its Agentic Storefronts.

Clicks only tell part of the story, though. In our shortlist study, 56 people completed shopping tasks in ChatGPT while our researchers tracked what they chose. Brands people chose had a 24% share of voice in ChatGPT's answers, compared with 11% for the brands they passed over, and 92.8% of tasks ended without a meaningful click to the open web.

The study concludes that, inside ChatGPT, share of voice works like shelf space. It measured choices in a test, not real purchases, but the point holds for measurement. Track your share of voice and shopping visibility alongside referral revenue, because referrals alone will undercount the channel.

Run AI shopping with your AI Marketer in Profound

Six jobs across a catalog of hundreds or thousands of products is more than a team can work through by hand. Someone has to decide which products need new pages, which listings are wrong, and which publication to pitch before the holidays. In Profound, AI Marketer (Aim) does that first pass for you.

Aim is the orchestrator inside Profound, a marketing teammate that reads everything Profound tracks for your brand. It works from Context Manager, where Profound stores six Brand Records: company, audience, competitors, offerings, strategy, and execution. Your team fills it from sources it already uses, like Slack, email, and Gong, so Aim knows your products, your positioning, and the claims you've approved.

Every week, Aim brings your team a ranked list of opportunities with the data behind each one, like a product line losing visibility, a sentiment drop around one product, or rising prompt volume in a category you don't own yet. When you approve an opportunity, Aim turns it into a project, a goal-based workspace with the plan and the tasks.

Say a competitor's products have started winning the comparison prompts in one of your categories. The project might include refreshing your best-sellers' pages with current specs, a category comparison page that answers the questions the competitor's page now answers, and a pitch for a gift guide the engines cite.

Aim builds and runs the agent for each task, and your team reviews every change before it's written to your store or sent to a publisher. It then tracks each task's impact on the project's metric and learns from what you accept, edit, or ignore.

See how your products show up in AI shopping today. Book a demo.

AI marketing agents for ecommerce FAQs

How do I get my products recommended in ChatGPT and Google AI Mode?

Give the engines complete, consistent product data in the places they read it, and earn mentions in the sources they trust. That means product feeds or catalog integrations like Shopify's Agentic Storefronts and Google Merchant Center, product pages that answer fit and use questions clearly, reviews that render in text, and coverage in the gift guides, roundups, and threads the engines cite.

Do I still need product pages if I have a product feed?

Yes. In our June 2026 analysis of ChatGPT Shopping, feeds supplied details like brand names and checkout images, while delivery information and availability came mostly from product pages. Feeds and pages fill in different parts of the answer.

Product pages also matter beyond ChatGPT. Personal agents like Muse browse websites directly, and Google's AI features draw on both Merchant Center data and the web. Keep pages and feeds consistent so every source gives the same price, availability, and specs.

Which product page changes matter most for AI shopping?

Start with an FAQ section that answers fit, sizing, compatibility, and use. In our Christmas in July study, top offers were 188% more likely to have one. Product videos, customer Q&A, reviews, and more images all showed up more often on top offers too.

Make sure the content is readable. Reviews that load through JavaScript may not be visible to answer engines, and clear product titles that mention common use cases help an engine match your product to a shopper's question. Stating who the product is for and its limitations helps both engines and browsing agents.

How do reviews affect AI product recommendations?

Reviews show up more often on the products AI ranks first, and shoppers use AI to make sense of them. In our Christmas in July study, top offers were 35% more likely to have customer reviews, and the IBM and NRF study found 33% of consumers use AI to interpret reviews.

Focus on volume, recency, and detail, and make sure reviews render in a format engines can read. Profound's Bazaarvoice integration can point review collection at the product attributes AI engines look for, so new reviews fill the most important gaps.

How do I measure revenue from AI shopping?

Combine referral data with visibility data. Use Agent Analytics and GA4 to see people arriving from answer engines and the revenue they bring, and check Shopify's ChatGPT referral attribution if you sell through its Agentic Storefronts.

Then add share of voice and shopping visibility for your key prompts, because many shoppers choose from AI's shortlist without clicking through. In our shortlist study, 92.8% of shopping tasks ended without a meaningful click, so referral revenue alone will understate what the channel does for you.