Agentic Shopping Is Under 1% of Etsy's Traffic

Four retailers described their AI spending inside the same week of August 2026, and the most useful sentence any of them said was a negative one. Etsy chief executive Kruti Goyal told Digital Commerce 360 that agentic AI experiences still account for less than 1% of Etsy's overall traffic as of the end of its fiscal second quarter. Not zero. Not a rounding error you can ignore forever. But less than one in a hundred visits, on a marketplace with about 87 million active buyers.

If you sell things and you have spent the past year reading that shopping agents are about to rewire your storefront, that number is worth sitting with. It comes from the company with the most to gain by talking the trend up, disclosed in the same breath as the parts of AI it actually paid for.

What did Etsy actually say about agentic shopping?

Goyal framed the company's AI work in three parts: making the marketplace more discoverable, making it more personal, and experimenting with what conversational commerce becomes. Only the third of those is the agent story, and it is the one she attached the smallest number to.

The detail sits in Etsy's own Q2 2026 shareholder letter filed with the SEC. The company launched a beta Gifting Assistant in May, described as a way for buyers to discover gift ideas through natural conversation. The letter says agent usage skews toward new, high intent buyers arriving from Etsy's own channels such as email, and that feedback has been positive. It does not claim volume. What the letter does quantify is unglamorous by comparison: richer buyer profiles now covering more than 65 million buyers with three times the signals they carried earlier in the year, averaging around 30 interests and preferences per buyer.

There is a second number in the letter that tells you where the engineering hours went. Etsy suppressed more than 80% of previously viewed and engaged listings from the buyer feed, and reports that this lifted favouriting, listing views and new searches while staying neutral on gross merchandise sales. That is a search relevance change, dressed in no futuristic language at all, and it is the kind of work that pays before an agent ever knocks.

Why does less than 1% still matter?

Because two different things are being counted, and merchants keep merging them. An agent completing a purchase is rare. A shopper consulting an assistant before they buy is not rare at all.

Pew Research Center surveyed 5,119 US adults between 17 and 23 February 2026 and found that about half of US adults now use AI chatbots, up from 33% in 2024, with 42% using them to search for information and 60% saying they at least sometimes read the AI summary at the top of a search results page. Those are the people already forming an opinion about your product before they see your site. Etsy's under 1% describes something narrower: the machine finishing the job on its own.

Etsy also observed that AI referral traffic shows higher purchase intent and higher average order value than other traffic sources, according to the Digital Commerce 360 account of Goyal's remarks. Small, and worth more per visit. That combination is exactly what a merchant should plan around: not a flood, but a trickle of people who arrive already convinced. We wrote about the mechanics of getting named in those answers in our piece on what actually moves your citation rate in AI answers, and nothing in this week's disclosures contradicts it.

Four disclosures, side by side

Four companies, one week, four different answers about where the money went. Put side by side, the pattern is that the disclosed numbers all sit on the boring side of the ledger.

What is Stanley 1913 changing about its product pages?

Of everything disclosed this week, this is the part a one person shop can lift wholesale. Chief brand officer Kate Ridley told Modern Retail that adapting to AI search is a muscle memory problem rather than a hacking problem, and the tactics behind that sentence are ordinary editorial work.

The brand is building explicit links between a product feature and the benefit a buyer gets from it, rather than leaving the reader to infer it. It is adding FAQs at the product level, along with care instructions and usage guides. It is reorganising content around the situations people actually ask about, gifting and hydration and travel and hosting, instead of around its own product taxonomy. It is leaning on written copy where it used to lean on imagery, because an image tells a language model nothing. And it is keeping structured data consistent so the same facts describe the product wherever they surface.

Notice what is absent. No new platform. No agent. No budget line a small merchant cannot match. The one piece that does cost money is measurement: Stanley told Modern Retail it uses Yotpo's Discovery product to track how often it appears in language model answers, and it watches the third party sources those systems treat as authorities, meaning earned media, reviews and affiliate coverage.

Do the Polywood numbers transfer to a smaller catalogue?

Chief digital officer Ben Spiegel gave Digital Commerce 360 the most quotable figures of the week: a 22% conversion rate increase and a 12% rise in average order value at the outdoor furniture retailer, alongside a claim that his developers ship twice as fast with nobody laid off. Read the rest of the account and the picture gets more specific, and less transferable.

Polywood renders lifestyle and studio imagery across 150,000 SKUs. It built customer personas around house types rather than personality types. Its predictive model leans on weather data, home sales records and real estate listings, and the company found that the seven day forecast was the most indicative signal for when somebody buys outdoor furniture. That is a genuinely clever finding. It is also the product of a category where the weather decides the purchase, applied to a catalogue with enough history for a model to learn from.

If you sell forty products, the equivalent of that project is not a model. It is looking at last year's sales next to last year's weather in a spreadsheet, which is the unglamorous starting point we set out in our guide to what a forecast needs before it can be useful. The 22% is real. The path to it is not a path most shops are standing on.

How should you read a 22% conversion claim?

Every one of these figures arrived without a control group. That is not an accusation, it is how retail communication works: a company reports what changed after it did something, and the sentence lands as cause and effect. When a vendor quotes one of these numbers back at you in a sales call, three questions separate a real result from a coincidence.

Ask what else changed in the same window. Polywood moved to a different ecommerce platform roughly eighteen months before the conversion figures it reported, and a replatform on its own moves conversion. Ask what the baseline was, because a lift from a low base and a lift from a strong base are different products. And ask which part of the funnel the number describes, since a conversion rate measured on sessions that reached a product page excludes everything that went wrong earlier.

None of that makes the 22% false. It makes it unquotable as a forecast for your shop. The useful part of the Polywood account is not the percentage, it is the mechanism: they found a signal outside their own data, the seven day weather forecast, that predicted demand better than their sales history did. Looking outside your own numbers for the thing that actually drives your season is advice worth taking whatever your catalogue size.

The distributor disclosure, and what it leaves out

US Foods spent part of its results call on digital and AI capability while reporting net sales up 4.5% year over year to $10.5 billion for a quarter ending 27 June. Nothing in that account gives a merchant an action, and it is included here for one reason: it is the shape most AI disclosure takes. A large company reports a growth number and an AI intention in the same paragraph, and the reader is left to connect them.

Set it against the Etsy letter and the difference is instructive. Etsy published a figure that made its agent programme look small, alongside figures that made its data work look substantial. That is a company telling you which of its bets has paid off. Most disclosures do not do that, which is why the ones that do are worth more than their headline.

What does a chief AI officer actually signal?

Target named Chandu Nair as its first senior vice president and chief AI officer, starting 24 August 2026, with Purvi Shah moving into a senior vice president role for user experience. Nair arrives from more than six years at Lowe's, most recently running stores, data, AI and innovation there.

The line worth keeping from the announcement is Nair's own framing: that the measure of success is the difference AI makes to growth, customer experience and employee efficiency, rather than how much of it gets deployed. That is a sentence written by somebody who has watched deployment counts become the metric. For a merchant, the signal in the appointment is not about tooling. It is that the largest retailers now think AI needs an owner with a seat, which usually happens after a couple of years of uncoordinated pilots.

Where does this leave a shop with no AI budget?

The honest summary of the week is that the biggest retailers are spending on being findable, and treating agents as a research project. A small merchant can follow the first half of that at close to zero cost, and should ignore the second half until the numbers move.

Start with the pages that already get traffic. For each one, ask whether a stranger could read it and state, in a sentence, who the product is for and why it beats the obvious alternative. If they could not, an assistant summarising your category cannot either. Add the questions customers actually email you, verbatim, as headings with answers underneath. Write the care and usage detail that lives in your head. Keep your product facts consistent between your site, your marketplace listings and your feeds, because inconsistency is what makes a model drop you rather than risk being wrong.

Then measure the one thing that matters: whether AI referrals are arriving and what they do when they land. Most analytics tools now separate that traffic. Adobe's analysis of over a trillion retail visits found AI referred shoppers converting 42% better than other channels in March 2026, which is why the segment repays reading separately even when it is tiny. If it is small and converts well, you have confirmation of the same pattern Etsy described at a scale you can see with your own eyes.

The agents, and what preparing for them costs

They are coming, slowly, and the preparation for them overlaps almost entirely with the preparation for AI search. An agent that cannot parse your product data cannot buy from you, which is the same failure as an assistant that cannot summarise your product page. The difference is what happens when the agent gets it wrong, and that is a question of access rather than of copy. We covered the liability side of that in our report on what a court decided about agents shopping on your site, and the advertising side in the piece on ads moving into the AI shopping assistant.

Stanley is testing catalogue integration into chat conversations through Shopify and Google's Universal Commerce Protocol. That is the right posture for a brand of its size: run the pilot, learn the format, keep the budget on the part that works today. A shop with one operator does not need to run that pilot. If protocols for feeding catalogues into chat settle into a standard, the platform underneath your storefront will implement it, which is one of the arguments for building on infrastructure that keeps up rather than assembling it yourself. Our own approach to building a commerce site is built on that assumption.

The number to carry into the holiday season

Less than 1%. That is what a marketplace with 87 million buyers, $2.6 billion in quarterly gross merchandise sales and every commercial reason to inflate the figure reported for agent driven traffic. Meanwhile 42% of US adults are already asking a chatbot questions, and Etsy says the visitors who arrive through AI referrals spend more than the ones who do not.

The gap between those two numbers is the whole opportunity for the next two quarters. It is not a technology gap. It is a writing gap, and the retailers with the biggest budgets spent this week telling everyone so.