0.4% of the traffic. 4% of the orders.
A US jewellery brand connected its Shopify store to AgentIQ in April 2026. Four months later: 5,425 product-quality fixes, 620 Agentic Product Listings live, 220 of them ranking top-10 — and AI agents driving 3.99% of the store's orders off 0.41% of its sessions.
Four numbers that carry the story
The first three come from the AgentIQ event log, Apr 7 – Aug 15, 2026. The fourth comes from store analytics, Jun 1 – Aug 16 — a shorter window, opening the month the first Agentic Product Listing went live. The page keeps the two apart throughout. Absolute counts and amounts are scaled by a constant factor to keep the merchant's trading volume confidential; every share, multiple and growth rate is the real one.
A beautiful catalog that agents could not read
The brand sells handcrafted sterling silver and gold jewellery — rings, necklaces, earrings and anklets, with most designs offered across several stones, metals and sizes. A few thousand SKUs, most of them a variation on a theme. That is exactly the catalog shape a human shopper browses happily and an AI shopping agent chokes on.
When a shopper asks an agent for "a personalised silver necklace for my mum's birthday, under $100", the agent is not reading the photography. It is reading structured data: which stone, which metal, which size, which occasion, which price band. If those attributes are buried in prose or missing altogether, the agent cannot build confidence in the match — so it recommends someone else.
On April 7, 2026 the store connected to AgentIQ and ran its first Catalog Audit. The opening score: 62 out of 100. Grade D.
How they adopted it, phase by phase
Every item below is a recorded AgentIQ event, in the order it happened. Volumes are scaled; see the note at the foot of the page.
Audit first, then fix the whole catalog at once
The first move was not optimisation, it was measurement. AgentIQ scored every product listing, found the failures, and pushed a bulk remediation pass before anyone touched a product page by hand.
- Apr 7First quality signals land. The first necklace listing reaches Excellent at 81/100.
- Apr 102,055 product listings improved in a single pass — all reach Excellent, 515 leave Bad status.
- Apr 10Catalog health 62 → 95. The catalog is upgraded from Grade D to Grade A.
Build the Prompt Library around real buying language
In parallel, the team seeded roughly 35 tracked prompts — not keywords, but the full sentences shoppers type at an agent. Each one arrived with an alignment score showing how well the catalog already answered it, which turned prompt selection into a prioritised worklist.
- Apr 7First 10 prompts tracked, including "show me sterling silver rings with a natural stone" at 60/100 and "find a personalised necklace for a birthday gift" at 80/100.
- Apr 8Twenty more added in one morning — category prompts (everyday necklaces 85, stacking ring sets 90) and gift prompts ("silver jewellery gifts under $100" 90).
- Apr 8Alignment on the natural-stone ring prompt improves 70 → 80 after remediation. The loop closes.
- Jun 23Prompt Library expansion: 30 new prompts generated from a single seed term across three sources.
Watch who else the agents are recommending
Competitor Data ran against the tracked prompts continuously. Across the period, 29 distinct competing stores were flagged as appearing against the brand's prompts — most of them repeatedly, on the single highest-value prompt in the library.
- Apr 7First competitor spotted at position 7 on the natural-stone ring prompt.
- Apr 11 → May 11The highest-value gift prompt alone produces six separate five-competitor alerts, with a rotating cast of direct-to-consumer jewellery labels and larger marketplace sellers.
- Apr 24A category prompt draws a completely different competitive set — a signal that two prompts can be separate battlegrounds rather than one ranking problem.
Grind the long tail up, automatically
This is the unglamorous middle of the story, and the part that made everything after it work. Thirty-four remediation passes over eleven weeks, with no manual schema rewriting on the merchant's side.
- Apr 17 – 18385 listings improved, then 460 reach Excellent. Health 90 → 94, Grade B → Grade A.
- Apr 20 – 21190 then 285 listings reach Excellent in back-to-back days.
- May 20Catalog health peaks at 97 / 100.
- throughoutIndividual listings climb on their own: a gemstone ring to 100/100, a pendant necklace 80 → 100, a charm necklace 97 → 100.
Publish Agentic Product Listings at scale
With the data layer clean and the brand's identity extracted from the storefront, AgentIQ started publishing Agentic Product Listings. The rollout was deliberately collection-led: the highest-volume core lines first, then the gemstone ranges, then the stacking and layering pieces.
- Jun 6Brand identity ready — voice, values, audience and buyer personas extracted from the storefront, refreshed on Aug 10.
- Jun 1195 listings go live in a single 70-minute window.
- Jun 22 – 25The peak burst: 260 listings published across four days, spanning four separate collections.
- Jul – AugSteady maintenance cadence. New products get an Agentic Product Listing within hours of analysis completing.
Rank in the Shopify Catalog
Two hundred and twenty product listings reached the top 10 catalog results for their prompts — roughly 35% of everything published, and still accruing at the end of the window.
- Jun 15First two rankings: an earring listing (top 10 for 3 prompts) and a stacking ring (2 prompts).
- Jun 22 & 25160 listings rank across two days — the June publishing burst converting in bulk.
- Aug 13 – 15All 30 of the most recent listings rank, each in the top 10 for 3 prompts.
Publishing and ranking, month by month
| Month | Published | Ranking |
|---|---|---|
| June 2026 | 530 | 175 |
| July 2026 | 60 | 5 |
| August 2026 | 30 | 40 |
Where the orders actually came from
Rankings are a leading indicator. Order volume is the one that pays. Across 77 days, AI agents accounted for 0.41% of the store's sessions and 3.99% of its orders — 10× the conversion their traffic share would predict. Hover any chart for a single day's figures.
| Day | Agentic sessions |
|---|---|
| Jun 1 | 40 |
| Jun 2 | 20 |
| Jun 3 | 15 |
| Jun 4 | 30 |
| Jun 5 | 25 |
| Jun 6 | 35 |
| Jun 7 | 35 |
| Jun 8 | 25 |
| Jun 9 | 35 |
| Jun 10 | 45 |
| Jun 11 | 20 |
| Jun 12 | 30 |
| Jun 13 | 30 |
| Jun 14 | 45 |
| Jun 15 | 50 |
| Jun 16 | 35 |
| Jun 17 | 35 |
| Jun 18 | 40 |
| Jun 19 | 30 |
| Jun 20 | 40 |
| Jun 21 | 30 |
| Jun 22 | 30 |
| Jun 23 | 25 |
| Jun 24 | 45 |
| Jun 25 | 10 |
| Jun 26 | 25 |
| Jun 27 | 15 |
| Jun 28 | 10 |
| Jun 29 | 10 |
| Jun 30 | 25 |
| Jul 1 | 25 |
| Jul 2 | 25 |
| Jul 3 | 25 |
| Jul 4 | 50 |
| Jul 5 | 40 |
| Jul 6 | 35 |
| Jul 7 | 35 |
| Jul 8 | 30 |
| Jul 9 | 25 |
| Jul 10 | 40 |
| Jul 11 | 55 |
| Jul 12 | 30 |
| Jul 13 | 35 |
| Jul 14 | 60 |
| Jul 15 | 35 |
| Jul 16 | 65 |
| Jul 17 | 40 |
| Jul 18 | 50 |
| Jul 19 | 35 |
| Jul 20 | 60 |
| Jul 21 | 35 |
| Jul 22 | 25 |
| Jul 23 | 65 |
| Jul 24 | 50 |
| Jul 25 | 50 |
| Jul 26 | 25 |
| Jul 27 | 55 |
| Jul 28 | 90 |
| Jul 29 | 50 |
| Jul 30 | 45 |
| Jul 31 | 40 |
| Aug 1 | 60 |
| Aug 2 | 55 |
| Aug 3 | 30 |
| Aug 4 | 60 |
| Aug 5 | 70 |
| Aug 6 | 55 |
| Aug 7 | 35 |
| Aug 8 | 50 |
| Aug 9 | 90 |
| Aug 10 | 50 |
| Aug 11 | 40 |
| Aug 12 | 95 |
| Aug 13 | 60 |
| Aug 14 | 65 |
| Aug 15 | 50 |
| Aug 16 | 85 |
Agentic orders per day, by how they arrived
| Day | ChatGPT-referred · 45 orders / $4,959 | Shop instant checkout · 80 orders / $5,646 |
|---|---|---|
| Jun 1 | 0 | 0 |
| Jun 2 | 0 | 0 |
| Jun 3 | 0 | 0 |
| Jun 4 | 0 | 5 |
| Jun 5 | 0 | 0 |
| Jun 6 | 0 | 0 |
| Jun 7 | 0 | 0 |
| Jun 8 | 0 | 0 |
| Jun 9 | 0 | 0 |
| Jun 10 | 0 | 0 |
| Jun 11 | 0 | 0 |
| Jun 12 | 0 | 0 |
| Jun 13 | 0 | 0 |
| Jun 14 | 0 | 0 |
| Jun 15 | 0 | 0 |
| Jun 16 | 0 | 0 |
| Jun 17 | 5 | 0 |
| Jun 18 | 0 | 5 |
| Jun 19 | 0 | 0 |
| Jun 20 | 0 | 5 |
| Jun 21 | 0 | 0 |
| Jun 22 | 0 | 0 |
| Jun 23 | 0 | 0 |
| Jun 24 | 0 | 0 |
| Jun 25 | 0 | 0 |
| Jun 26 | 0 | 0 |
| Jun 27 | 0 | 5 |
| Jun 28 | 0 | 0 |
| Jun 29 | 0 | 0 |
| Jun 30 | 0 | 0 |
| Jul 1 | 0 | 5 |
| Jul 2 | 0 | 0 |
| Jul 3 | 0 | 0 |
| Jul 4 | 5 | 0 |
| Jul 5 | 0 | 0 |
| Jul 6 | 0 | 0 |
| Jul 7 | 0 | 0 |
| Jul 8 | 0 | 0 |
| Jul 9 | 0 | 0 |
| Jul 10 | 0 | 0 |
| Jul 11 | 5 | 0 |
| Jul 12 | 0 | 10 |
| Jul 13 | 0 | 0 |
| Jul 14 | 0 | 0 |
| Jul 15 | 0 | 0 |
| Jul 16 | 0 | 0 |
| Jul 17 | 0 | 0 |
| Jul 18 | 0 | 0 |
| Jul 19 | 0 | 0 |
| Jul 20 | 5 | 0 |
| Jul 21 | 0 | 0 |
| Jul 22 | 0 | 0 |
| Jul 23 | 0 | 0 |
| Jul 24 | 5 | 0 |
| Jul 25 | 0 | 0 |
| Jul 26 | 0 | 0 |
| Jul 27 | 0 | 0 |
| Jul 28 | 0 | 0 |
| Jul 29 | 0 | 0 |
| Jul 30 | 0 | 0 |
| Jul 31 | 0 | 0 |
| Aug 1 | 0 | 10 |
| Aug 2 | 0 | 0 |
| Aug 3 | 0 | 5 |
| Aug 4 | 0 | 0 |
| Aug 5 | 0 | 10 |
| Aug 6 | 0 | 0 |
| Aug 7 | 0 | 5 |
| Aug 8 | 0 | 0 |
| Aug 9 | 10 | 0 |
| Aug 10 | 0 | 0 |
| Aug 11 | 0 | 0 |
| Aug 12 | 5 | 5 |
| Aug 13 | 5 | 0 |
| Aug 14 | 0 | 5 |
| Aug 15 | 0 | 0 |
| Aug 16 | 0 | 5 |
Agentic share of orders, 7-day rolling
| Day | Agentic share of orders |
|---|---|
| Jun 1 | 0.00% |
| Jun 2 | 0.00% |
| Jun 3 | 0.00% |
| Jun 4 | 4.3% |
| Jun 5 | 3.4% |
| Jun 6 | 2.4% |
| Jun 7 | 2.0% |
| Jun 8 | 2.2% |
| Jun 9 | 1.85% |
| Jun 10 | 1.67% |
| Jun 11 | 0.00% |
| Jun 12 | 0.00% |
| Jun 13 | 0.00% |
| Jun 14 | 0.00% |
| Jun 15 | 0.00% |
| Jun 16 | 0.00% |
| Jun 17 | 2.4% |
| Jun 18 | 4.5% |
| Jun 19 | 4.5% |
| Jun 20 | 6.3% |
| Jun 21 | 5.8% |
| Jun 22 | 4.5% |
| Jun 23 | 4.3% |
| Jun 24 | 2.6% |
| Jun 25 | 1.35% |
| Jun 26 | 1.35% |
| Jun 27 | 1.35% |
| Jun 28 | 1.39% |
| Jun 29 | 1.67% |
| Jun 30 | 1.59% |
| Jul 1 | 3.4% |
| Jul 2 | 3.5% |
| Jul 3 | 3.6% |
| Jul 4 | 3.9% |
| Jul 5 | 3.9% |
| Jul 6 | 4.3% |
| Jul 7 | 4.9% |
| Jul 8 | 2.4% |
| Jul 9 | 2.6% |
| Jul 10 | 2.5% |
| Jul 11 | 2.4% |
| Jul 12 | 7.0% |
| Jul 13 | 6.1% |
| Jul 14 | 6.0% |
| Jul 15 | 5.7% |
| Jul 16 | 5.5% |
| Jul 17 | 5.6% |
| Jul 18 | 3.7% |
| Jul 19 | 0.00% |
| Jul 20 | 1.85% |
| Jul 21 | 2.0% |
| Jul 22 | 2.2% |
| Jul 23 | 2.1% |
| Jul 24 | 4.0% |
| Jul 25 | 3.8% |
| Jul 26 | 4.3% |
| Jul 27 | 2.2% |
| Jul 28 | 1.85% |
| Jul 29 | 1.96% |
| Jul 30 | 1.82% |
| Jul 31 | 0.00% |
| Aug 1 | 4.2% |
| Aug 2 | 4.0% |
| Aug 3 | 5.5% |
| Aug 4 | 5.8% |
| Aug 5 | 7.9% |
| Aug 6 | 9.1% |
| Aug 7 | 8.3% |
| Aug 8 | 4.9% |
| Aug 9 | 7.0% |
| Aug 10 | 6.2% |
| Aug 11 | 6.1% |
| Aug 12 | 6.3% |
| Aug 13 | 6.6% |
| Aug 14 | 8.0% |
| Aug 15 | 8.7% |
| Aug 16 | 6.8% |
Six things this data says
Agents drove 3.99% of orders off 0.41% of sessions. 80 of the 125 agentic orders closed inside Shop's instant checkout with no storefront session at all; the other 45 were credited to ChatGPT clicks through cross-journey attribution. Session counts on their own dramatically undersell this channel — which is precisely why a traffic dashboard makes it look like nothing is happening.
This is a ChatGPT story. 3,135 of 3,160 agentic sessions came from ChatGPT — no Copilot, no Perplexity, and only 25 Shop-referred sessions across 77 days. For a merchant deciding where to spend attention, the agentic channel is currently one surface, and it is answering shoppers' jewellery questions one conversation at a time.
Agent-referred buyers spend more than average. $110 per order against a $96 store AOV, about 15% higher. Shop instant-checkout orders average $71 — the quick-buy pattern. Shoppers arriving from an agent are pre-qualified: they described what they wanted before they ever clicked.
ChatGPT cites content, not only products. A category buying guide is a top-five agentic landing page, sitting alongside a long tail of individual product pages. Answer-shaped editorial is an acquisition surface in its own right, and it is the part of the catalog most merchants never think to optimise for agents.
The SEO foundation is doing the agentic work. Roughly 305k of 774k sessions come from organic search with essentially no paid traffic. The same crawlable, well-ranked catalog that wins Google is what an agent retrieves and recommends — agentic visibility is not a separate channel you buy, it is a property of the data layer you already own.
The honest caveat. Only 25 agentic sessions completed checkout on-site — about 0.8%. The rest of the value closes later, or inside Shop, and is credited to an agent click from an earlier visit rather than a same-day session. That is normal here and it is why session-level conversion looks poor while order share looks excellent. Both numbers are real; only one of them is the point.
Top agentic landing pages
Agentic sessions by landing page. Pages are described by type rather than named, to keep the merchant anonymous. The homepage leads, but the long tail of individual product pages is where the agentic intent is most specific.
Daily detail, all 77 days
Highlighted rows are days with at least one agentic order.
| Day | Sessions | Agentic | Share | Orders | Gross | Agentic orders | Agentic gross |
|---|---|---|---|---|---|---|---|
| Jun 1 | 10,280 | 40 | 0.39% | 35 | $4,463 | – | – |
| Jun 2 | 13,895 | 20 | 0.14% | 10 | $859 | – | – |
| Jun 3 | 6,260 | 15 | 0.24% | 20 | $1,105 | – | – |
| Jun 4 | 6,415 | 30 | 0.47% | 50 | $3,600 | 5 | $114 |
| Jun 5 | 7,240 | 25 | 0.35% | 30 | $2,214 | – | – |
| Jun 6 | 11,075 | 35 | 0.32% | 65 | $5,171 | – | – |
| Jun 7 | 11,410 | 35 | 0.31% | 40 | $3,577 | – | – |
| Jun 8 | 10,310 | 25 | 0.24% | 15 | $1,608 | – | – |
| Jun 9 | 11,180 | 35 | 0.31% | 50 | $4,166 | – | – |
| Jun 10 | 11,930 | 45 | 0.38% | 50 | $5,974 | – | – |
| Jun 11 | 17,895 | 20 | 0.11% | 45 | $4,133 | – | – |
| Jun 12 | 13,745 | 30 | 0.22% | 30 | $2,288 | – | – |
| Jun 13 | 11,540 | 30 | 0.26% | 25 | $2,291 | – | – |
| Jun 14 | 12,770 | 45 | 0.35% | 25 | $2,298 | – | – |
| Jun 15 | 10,390 | 50 | 0.48% | 30 | $3,301 | – | – |
| Jun 16 | 10,190 | 35 | 0.34% | 30 | $3,570 | – | – |
| Jun 17 | 8,695 | 35 | 0.40% | 25 | $2,669 | 5 | $586 |
| Jun 18 | 12,080 | 40 | 0.33% | 55 | $7,456 | 5 | $325 |
| Jun 19 | 17,175 | 30 | 0.17% | 30 | $2,425 | – | – |
| Jun 20 | 9,990 | 40 | 0.40% | 45 | $3,978 | 5 | $215 |
| Jun 21 | 14,705 | 30 | 0.20% | 45 | $4,905 | – | – |
| Jun 22 | 10,250 | 30 | 0.29% | 100 | $12,738 | – | – |
| Jun 23 | 7,645 | 25 | 0.33% | 50 | $4,417 | – | – |
| Jun 24 | 8,625 | 45 | 0.52% | 60 | $5,291 | – | – |
| Jun 25 | 24,150 | 10 | 0.04% | 40 | $4,573 | – | – |
| Jun 26 | 9,795 | 25 | 0.26% | 30 | $3,178 | – | – |
| Jun 27 | 12,660 | 15 | 0.12% | 45 | $3,798 | 5 | $223 |
| Jun 28 | 18,155 | 10 | 0.06% | 35 | $2,915 | – | – |
| Jun 29 | 11,980 | 10 | 0.08% | 40 | $5,764 | – | – |
| Jun 30 | 11,520 | 25 | 0.22% | 65 | $6,540 | – | – |
| Jul 1 | 10,450 | 25 | 0.24% | 35 | $4,347 | 5 | $177 |
| Jul 2 | 9,165 | 25 | 0.27% | 35 | $2,607 | – | – |
| Jul 3 | 14,705 | 25 | 0.17% | 25 | $1,886 | – | – |
| Jul 4 | 8,785 | 50 | 0.57% | 20 | $1,677 | 5 | $609 |
| Jul 5 | 12,610 | 40 | 0.32% | 35 | $3,156 | – | – |
| Jul 6 | 8,095 | 35 | 0.43% | 15 | $1,009 | – | – |
| Jul 7 | 10,060 | 35 | 0.35% | 40 | $3,448 | – | – |
| Jul 8 | 7,965 | 30 | 0.38% | 35 | $3,700 | – | – |
| Jul 9 | 6,545 | 25 | 0.38% | 20 | $1,889 | – | – |
| Jul 10 | 7,115 | 40 | 0.56% | 35 | $2,064 | – | – |
| Jul 11 | 10,665 | 55 | 0.52% | 25 | $2,634 | 5 | $725 |
| Jul 12 | 8,295 | 30 | 0.36% | 45 | $4,150 | 10 | $1,004 |
| Jul 13 | 14,975 | 35 | 0.23% | 45 | $4,228 | – | – |
| Jul 14 | 8,945 | 60 | 0.67% | 45 | $3,658 | – | – |
| Jul 15 | 11,175 | 35 | 0.31% | 50 | $4,971 | – | – |
| Jul 16 | 8,415 | 65 | 0.77% | 30 | $1,787 | – | – |
| Jul 17 | 7,965 | 40 | 0.50% | 30 | $4,183 | – | – |
| Jul 18 | 6,240 | 50 | 0.80% | 25 | $1,924 | – | – |
| Jul 19 | 11,210 | 35 | 0.31% | 45 | $3,793 | – | – |
| Jul 20 | 8,570 | 60 | 0.70% | 45 | $2,995 | 5 | $375 |
| Jul 21 | 6,890 | 35 | 0.51% | 20 | $1,328 | – | – |
| Jul 22 | 6,665 | 25 | 0.38% | 35 | $3,542 | – | – |
| Jul 23 | 8,190 | 65 | 0.79% | 35 | $4,352 | – | – |
| Jul 24 | 8,870 | 50 | 0.56% | 45 | $4,842 | 5 | $614 |
| Jul 25 | 6,430 | 50 | 0.78% | 35 | $3,453 | – | – |
| Jul 26 | 7,465 | 25 | 0.33% | 20 | $1,423 | – | – |
| Jul 27 | 7,975 | 55 | 0.69% | 40 | $4,473 | – | – |
| Jul 28 | 10,440 | 90 | 0.86% | 60 | $4,149 | – | – |
| Jul 29 | 7,390 | 50 | 0.68% | 20 | $2,346 | – | – |
| Jul 30 | 7,415 | 45 | 0.61% | 55 | $7,844 | – | – |
| Jul 31 | 6,585 | 40 | 0.61% | 15 | $1,883 | – | – |
| Aug 1 | 6,120 | 60 | 0.98% | 30 | $2,388 | 10 | $1,129 |
| Aug 2 | 6,900 | 55 | 0.80% | 30 | $1,432 | – | – |
| Aug 3 | 17,155 | 30 | 0.17% | 65 | $5,903 | 5 | $625 |
| Aug 4 | 7,545 | 60 | 0.80% | 45 | $3,272 | – | – |
| Aug 5 | 9,075 | 70 | 0.77% | 75 | $7,849 | 10 | $699 |
| Aug 6 | 12,735 | 55 | 0.43% | 15 | $577 | – | – |
| Aug 7 | 10,950 | 35 | 0.32% | 100 | $10,138 | 5 | $205 |
| Aug 8 | 7,980 | 50 | 0.63% | 80 | $6,021 | – | – |
| Aug 9 | 7,645 | 90 | 1.18% | 50 | $4,038 | 10 | $1,146 |
| Aug 10 | 8,250 | 50 | 0.61% | 40 | $3,734 | – | – |
| Aug 11 | 8,215 | 40 | 0.49% | 50 | $5,955 | – | – |
| Aug 12 | 9,650 | 95 | 0.98% | 65 | $5,627 | 10 | $768 |
| Aug 13 | 8,485 | 60 | 0.71% | 70 | $6,974 | 5 | $359 |
| Aug 14 | 8,340 | 65 | 0.78% | 20 | $2,133 | 5 | $409 |
| Aug 15 | 7,130 | 50 | 0.70% | 50 | $5,859 | – | – |
| Aug 16 | 7,475 | 85 | 1.14% | 70 | $7,638 | 5 | $300 |
Pulled 2026-08-16 from AgentIQ analytics for this store. Agentic sessions are those the store attributes to an AI agent referral. Agentic orders are those attributed to an agentic referral or an agentic sales channel, which is what folds Shop instant-checkout orders in and keeps the totals aligned with the agentic channel as Shopify admin reports it. Because an order can be credited to an agent click from an earlier visit, a day's orders do not require that day's sessions — which is why the two series do not track row for row, and why session share alone understates the channel. Gross sales are in USD, on the store's local clock. August 16 is a partial day. Counts and amounts are scaled by a constant factor; shares and averages are unaffected.
The share of orders more than tripled in ten weeks
The number that matters for a return is not the total, it is the slope. Agentic order share roughly tripled across the window while store-wide sessions cooled from their June peak — the channel grew on both ends at once.
$10,574 a month, and rising
August produced $5,639 in its first 16 days — a pace of about $10,574 a month against $1,462 in the whole of June. The trajectory, not the $10,605 total, is what a merchant is buying.
No paid spend behind any of it
Roughly 39% of store traffic is organic search and there is essentially no paid traffic. The agentic channel was not bought — it came from making the existing catalog legible. There is no media cost to net off against the return.
A higher-value order, too
Agent-referred orders average $110 against a $96 store AOV, about 15% higher. The channel is not just incremental volume at the same margin; the shoppers arriving through it had already described what they wanted.
This is attributed revenue, not proven incremental revenue. Some of these 125 shoppers might have found the store another way, and nothing here is a holdout test. Read it as evidence that a channel exists and is compounding, not as a lift measurement.
Subscription cost is also deliberately absent: the plan tier is not in either dataset, so the page reports return as revenue and pace rather than as a ratio it cannot substantiate.
And the dollar figures are scaled, not this merchant's actual takings — so read the pace as a shape and the percentages as fact.
Everything, from both sources
Two systems, two windows, kept apart on purpose. The catalog figures come from the AgentIQ event log (Apr 7 – Aug 15); the traffic and revenue figures come from store analytics (Jun 1 – Aug 16). Absolute values are scaled; shares, averages and multiples are the real ones.
Catalog and rankings · AgentIQ event log
| Metric | Value | Evidence |
|---|---|---|
| Observation window | Apr 7 – Aug 15, 2026 | 131 days, first quality event to latest ranking event |
| Opening catalog health | 62 / 100 · Grade D | Baseline before the first remediation pass |
| Catalog health after the first bulk pass | 95 / 100 · Grade A | Apr 10, three days after connecting |
| Peak catalog health | 97 / 100 · Grade A | May 20, 2026 |
| Grade A upgrades recorded | 2 | Apr 10 (from D) and Apr 18 (from B) |
| Product-quality improvements | 5,425 | Sum of 34 recorded remediation events |
| Largest single pass | 2,055 listings | Apr 10; 515 listings left Bad status |
| Listings scoring 100 / 100 | 85+ | Gemstone rings, pendant necklaces, sterling-silver chains |
| Tracked prompts in the Prompt Library | ~35 | Alignment scores 0–90 at creation |
| Prompt Library expansion | 30 prompts | Jun 23, one seed term, three sources |
| Competing stores identified | 29 | Across 11 competitor-detection events |
| Agentic Product Listings published | 620 | Distinct listings, Jun 1 – Aug 12 |
| Listings ranking in the Shopify Catalog | 220 | Top 10 catalog results for 2–3 prompts each |
| Publish-to-rank conversion | 35% | 220 of 620, and still accruing |
| Brand identity extractions | 2 | Jun 6 initial, Aug 10 refresh |
| Manual schema rewriting by the merchant | 0 | No manual optimisation events recorded in the log |
Traffic and revenue · store analytics
| Metric | Value | Evidence |
|---|---|---|
| Traffic window | Jun 1 – Aug 16, 2026 | 77 days, Aug 16 partial · store local time · USD, scaled |
| Store sessions | 773,865 | ~10,050 per day |
| Agentic sessions | 3,160 | 0.41% of sessions, present on all 77 days |
| Agentic sessions by source | 3,135 / 25 | ChatGPT / Shop. No Copilot or Perplexity sessions recorded |
| Store orders | 3,130 | $96.02 average order value |
| Agentic orders | 125 · 3.99% | 10× the session share |
| Agentic orders by route | 45 / 80 | ChatGPT-referred / closed in Shop instant checkout |
| Store gross sales | $300,538 | Same 77-day window |
| Agentic gross sales | $10,605 | 3.53% of store gross |
| Agentic gross by month | $1,462 → $3,503 → $5,639 | June, July, first 16 days of August |
| Agentic share of orders by month | 2.1% → 3.3% → 7.6% | More than tripled across the window |
| Agentic average order value | $84.84 | Blended. $110 referred, $71 Shop instant checkout |
| Peak agentic sessions in a day | 95 | Aug 12, 2026 |
| Days with zero agentic sessions | 0 | Agents visited on all 77 days |
| On-site agentic checkouts | 25 | ~0.8% of agentic sessions. The rest close later or in Shop |
Three lessons for the next merchant
Fix the catalog before you chase prompts
The brand spent its first eight weeks almost entirely on quality remediation and published nothing agentic until June. When publishing did start, 35% of listings ranked. A clean data layer is not a prerequisite you rush past — it is the thing that makes everything downstream convert.
Variation catalogs are the strongest fit
A dozen stones across a dozen designs in three metals and four sizes is combinatorially brutal for a human merchandiser and close to ideal for structured optimisation. Every attribute an agent needs — stone, metal, size, occasion, price band — already existed as a real product difference. It simply was not machine-readable.
Judge the channel on orders, not sessions
At 0.41% of sessions the agentic channel looks like a rounding error, and a traffic dashboard will tell you to ignore it. At 3.99% of orders — one in twenty-five — it is a channel. Two thirds of those orders never produced a storefront session at all, so any measurement that starts from visits will keep reporting that nothing is happening.
Two sources, never blended. The catalog, listing and ranking figures come from AgentIQ product events recorded between April 7 and August 15, 2026. The traffic and revenue figures come from an analytics pull dated August 16, 2026 covering June 1 – August 16 — a shorter window, opening the month the first Agentic Product Listing went live. Every figure on the page is labelled with the window it belongs to.
Counts of Agentic Product Listings and rankings are de-duplicated by listing. The product-quality total is the sum of per-event counts, so a listing improved across two passes counts twice. Attribution details for the commercial figures are in the traffic-data method note above.
The customer is disguised, and the absolute figures are scaled. The store name, domain, product titles, collections, competitor names and home market are all withheld or generalised, and every count and amount on this page is multiplied by a constant factor so the merchant's real trading volume stays confidential.
That transformation is linear, which is the point: every percentage, ratio, multiple, average order value and growth rate here is exactly the real one, because each is a quotient of two scaled numbers. Read the shares and the multiples as fact. Read the totals as correctly proportioned, not as this merchant's books.
this took one install and no manual rewriting.
AgentIQ audits your catalog, fixes what agents cannot read, and publishes Agentic Product Listings for you. Shopify-native, and the first audit runs on the day you connect.