--- title: ChosenByAgent — See where AI assistants send your buyers canonical: https://chosenbyagent.com/ updated: 2026-10-22 audience: both offers: - A5 - BP - A2 - B2 - B3 - A3 - B4 - A4 - A6 evidence: - chatgpt-session-share - agentic-checkout-share - api-consumer-overlap - non-shopify-stalls description: Independent testing of how AI assistants and shopping agents route demand, show your listings and get through your site. For marketplaces and brands. --- # See where AI assistants send your buyers — and fix what loses them. Independent AI-channel testing · for marketplaces and brands We run ChatGPT, Gemini, Claude and shopping agents through your real buyer questions. You see where they route demand, how they show your listings, where they fail on your site — and what to fix first. Reading · routing (Illustrative): “used road bike under €300, Lyon” — gpt-6.1-sol: vertical, gemini-3.8: vertical, sonnet-5.5: horizontal. Not client data · real audits run 15 repeats per cell. ## Small today, decided by a few models — and it shifts with every release. - **0.2%** — ChatGPT drove about 0.2% of e-commerce sessions across 973 stores between August 2024 and July 2025. [Marketing Science (INFORMS), 2025, independent](https://pubsonline.informs.org/doi/full/10.1287/mksc.2025.0489) - **3.16%** — 3.16% of agentic requests reach checkout and payment, while 69.6% land on product and search pages. [HUMAN, 2026-04, vendor](https://humansecurity.com/learn/blog/state-of-agentic-traffic-april-26) - **15–32%** — Brands named in API answers overlap with what consumer assistant apps show on only 15–32% of brands. [Surfer, 2026, vendor](https://surferseo.com/blog/llm-scraped-ai-answers-vs-api-results) - **79 / 123** — A shopping agent stalled on 79 of 123 stores outside Shopify, versus 5 of 121 Shopify stores. [agent-checkout-study (D. Walia), 2026, independent](https://github.com/d-walia/agent-checkout-study) ## Start with the question you’d ask an assistant. Each question, with its audience (Marketplaces, Brands or Both), the instrument that answers it and its price: - [Q01 · Where do AI assistants send my buyers?](/questions/where-do-assistants-send-buyers.md) — Marketplaces · A2 Routing & listing audit · from €2,500 (one-off) - [Q05 · Can a shopping agent actually get through my site?](/questions/can-an-agent-get-through.md) — Brands · B2 Agent access audit · from €1,500 (one-off) - [Q02 · Do assistants show my listings correctly?](/questions/are-my-listings-shown-right.md) — Marketplaces · A2 Routing & listing audit · from €2,500 (one-off) - [Q06 · Which brand does the agent pick — and why?](/questions/which-brand-does-the-agent-pick.md) — Brands · B3 Choice test · from €3,000 (per category) - [Q03 · Does my ChatGPT app or MCP work as intended?](/questions/does-my-ai-app-work.md) — Marketplaces · A3 AI channel check · from €3,000 (one-off) - [Q07 · What changes if I change price, returns or copy?](/questions/what-if-i-change-the-offer.md) — Brands · B4 Offer lab · from €1,000 (per experiment) - [Q04 · Do buyer and seller agents behave safely on my platform?](/questions/are-agents-safe-in-my-chats.md) — Marketplaces · A4 Agent red team · from €5,000 (one-off) - [Q08 · What changed after the last model update?](/questions/what-changed-after-the-update.md) — Both · A6 Monitoring · from €800 (per month) ## Pilots 4 weeks · fixed scope ### Marketplace pilot Marketplaces & classifieds. Routing & listing audit, AI channel check and agent red team, calibrated against your assistant referral traffic. A5 · from €7,500 (one-off) · A2 + A3 + A4 €10,500 separately · 4 weeks ### Brand pilot Brands & e-commerce. Agent access audit plus a choice test in one category — where agents stop, whom they pick and why. BP · from €4,000 (one-off) · B2 + B3 €4,500 separately · 4 weeks ## Measured honestly, or not at all. 1. **Run** — Real buyer intents in your markets and languages, through official APIs with web search, many repeats per cell. 2. **Stamp** — Model, harness version, date. Comparisons across a model update are flagged. 3. **Map** — Sold-as-live, wrong prices, lost attribution, blocked agents — checked by hand and against your logs. 4. **Fix** — Prioritised changes with an owner each, and a re-run after the next model release. **What we don’t claim:** forecasts of what any consumer agent will buy, or “agent conversion”. Our numbers are a versioned index. `RUN · gpt-6.1-sol · harness 0.4 · 2026-10-21` — [Read the method](/method.md) ## No fees from AI vendors, platforms or agencies we test. We are paid by the companies we measure for, never by the assistants, platforms or agencies whose behaviour we report. ## Free start ### Benchmark v0 For marketplaces. How assistants route demand across European classifieds: 5 markets, 4–5 languages, ~12k queries, 200 manual runs. In preparation — get early findings before publication. [Get early findings](/benchmark.md) ### Free agent snapshot For brands. One scenario on your store, three models, about 30 runs. See exactly where a shopping agent stops. [Get the snapshot](/questions/can-an-agent-get-through.md) ## We practise what we test. Agents can read every page as Markdown, compare our prices and request work for you. You confirm by email before anything happens. ```sh # read any page as Markdown curl -H "Accept: text/markdown" \ https://chosenbyagent.com/questions/can-an-agent-get-through # request work for a person (they confirm by email) curl -X POST https://chosenbyagent.com/api/engagements \ -H "Content-Type: application/json" \ -d '{"offer_id":"B2","on_behalf_of":{"email":"…"},"agent":{"name":"…"}}' → 202 pending_confirmation ``` Formats, limits and confirmation: /agents.md · /llms.txt · /catalog.json ## Tell us what you need read. We reply with a scope, a fixed price and a start date. Agents: POST /api/engagements (application/json) with {"offer_id": "A5", "on_behalf_of": {"email": "…"}, "agent": {"name": "…"}}. The person confirms by email before we see the request. Details: /agents.md