Agent Readiness Index 2027: web summary
How ready online stores were for AI shopping agents when we measured them: results by level, sector, platform and check, the method, the limits of the sample and the aggregated data as CSV. No store names, no form.
- Scope
- 836 online stores: 365 in Türkiye, 471 English-language
- Measured
- October 6, 2026, with the 12 checks of the first version (v1) of the free test
- Vantage points
- The same scan from a home connection and from cloud egress
- Store names
- None published; aggregated results only. No store pays to be included.
Source: Specoria, Agent Readiness Index 2027 (PDF) · · 836 stores
Download the PDF Türkiye edition (Turkish PDF) →
Results
Every figure below is printed in the PDF report; the page number is given under each table.
English-language edition (471 stores)
Key figures
- Average agent-readiness score 67/100 (median 71), across 343 stores with a product page found.
- Only 18 of 343 stores (5.2%) pass all six product-level checks.
- Shipping terms are machine-readable in 9.3% of stores, return terms in 12.0%.
- 42 of 163 well-known retailers and brands (26%) refused our automated homepage request; 0 of the other 308 stores did.
- Only 5 of 436 readable robots.txt files block a bot that shopping agents use.
Source: Specoria, Agent Readiness Index 2027 (PDF) · · PDF pages 5, 9, 15
Levels
Share of the 343 scored stores at each level, and how their scores are spread.
| Level | Score | Share |
|---|---|---|
| Hard for agents to read | score below 40 | 11% |
| Partly ready | score 40-69 | 35% |
| Looks ready | score 70 and above | 54% |
Score band (stores): 0-9: 0 · 10-19: 1 · 20-29: 8 · 30-39: 30 · 40-49: 45 · 50-59: 26 · 60-69: 48 · 70-79: 73 · 80-89: 90 · 90-100: 22
Source: Specoria, Agent Readiness Index 2027 (PDF) · · PDF page 5
The “fully ready” funnel
The 343 stores with a product page, kept step by step while they pass each product-level check.
| Step | Share | Stores |
|---|---|---|
| Product structured data present | 70% | 241 |
| … price, currency and stock complete | 66% | 227 |
| … brand and GTIN/MPN defined | 29% | 98 |
| … price in first HTML (no JavaScript) | 29% | 98 |
| … return policy machine-readable | 7% | 24 |
| … shipping info machine-readable | 5% | 18 |
Source: Specoria, Agent Readiness Index 2027 (PDF) · · PDF page 5
By sector
Average score and the share of stores passing each product-level check, per sector.
| Sector | Stores | With product page | Avg. score | Product data | Price / stock | Brand + GTIN | Shipping | Returns |
|---|---|---|---|---|---|---|---|---|
| Fashion and accessories | 103 | 70 | 72 | 79% | 77% | 41% | 14% | 19% |
| Beauty and personal care | 57 | 49 | 69 | 71% | 67% | 31% | 6% | 10% |
| Home and living | 62 | 39 | 68 | 77% | 69% | 26% | 10% | 10% |
| Baby, kids and toys | 32 | 27 | 67 | 70% | 67% | 22% | 7% | 11% |
| Sports and outdoor | 38 | 26 | 66 | 69% | 65% | 23% | 8% | 8% |
| Electronics | 69 | 49 | 64 | 63% | 57% | 29% | 12% | 14% |
| Food, drink and supplements | 43 | 33 | 63 | 70% | 70% | 27% | 6% | 9% |
| Other and specialty | 67 | 50 | 63 | 60% | 54% | 24% | 6% | 8% |
Sector assigned from the page title and description keywords (and from our list for well-known retailers). Sector samples are small and their platform mixes differ.
Source: Specoria, Agent Readiness Index 2027 (PDF) · · PDF page 10
By platform
Platform from the homepage fingerprint; the score and product columns use stores with a product page.
| Platform | Stores | Product page found | Avg. score | Product data | Price / stock | Brand + GTIN | Shipping | Returns | llms.txt |
|---|---|---|---|---|---|---|---|---|---|
| Salesforce Commerce Cloud | 13 | 11 | 79 | 100% | 91% | 73% | 27% | 18% | 8% |
| Shopify | 143 | 140 | 78 | 83% | 83% | 48% | 11% | 16% | 96% |
| WooCommerce | 56 | 48 | 61 | 75% | 69% | 0% | 0% | 2% | 19% |
| BigCommerce | 51 | 48 | 60 | 65% | 63% | 13% | 8% | 10% | 8% |
| Other / custom | 95 | 56 | 58 | 55% | 46% | 25% | 7% | 9% | 27% |
| Magento / Adobe Commerce | 52 | 33 | 56 | 45% | 36% | 18% | 15% | 15% | 4% |
Platform shares in the sample were set by us, so compare within a platform rather than reading the overall average as a market figure.
Source: Specoria, Agent Readiness Index 2027 (PDF) · · PDF pages 8, 16
Check by check
The 12 checks with their weight, the number of stores where each could be measured, the share that pass and the 95% interval (within the sample only).
| Check | Weight | Measured | Pass | 95% interval |
|---|---|---|---|---|
| AI agent access allowed (robots.txt) | 15 | 471 | 93.0% | 90-95 |
| Homepage reachable by automated requests | 10 | 471 | 89.4% | 86-92 |
| Price readable without JavaScript | 5 | 343 | 81.6% | 77-85 |
| Sitemap | 5 | 430 | 85.6% | 82-89 |
| Product structured data (schema.org) | 12 | 343 | 70.3% | 65-75 |
| Price, currency, stock | 10 | 343 | 66.2% | 61-71 |
| Product identity (brand + GTIN/MPN) | 10 | 343 | 29.4% | 25-35 |
| Machine-readable shipping info | 6 | 343 | 9.3% | 7-13 |
| Machine-readable return policy | 7 | 343 | 12.0% | 9-16 |
| Organization data | 6 | 421 | 58.7% | 54-63 |
| Return policy page | 8 | 421 | 60.1% | 55-65 |
| llms.txt file | 6 | 425 | 43.8% | 39-49 |
Source: Specoria, Agent Readiness Index 2027 (PDF) · · PDF page 15
Who shuts automated visitors out
Share of stores whose homepage answered our automated request with an access-denied status. A generic automated client was used, not real agent identities.
| Stores | Share | |
|---|---|---|
| Well-known retailers and brands, from the home connection | 42/163 | 26% |
| All other stores, from the home connection | 0/308 | 0% |
| All stores, from cloud egress | 39/469 | 8.3% |
Source: Specoria, Agent Readiness Index 2027 (PDF) · · PDF page 9
Türkiye edition (365 stores)
Not in the web summary yet
The Türkiye edition (365 stores) results are in its PDF report (in Turkish). They will appear here once they have been transferred from that PDF; until then this page shows no figures for it.
Download the data
Every figure on this page as a CSV (one row per figure, with its PDF page), licensed under Creative Commons Attribution 4.0 (CC BY 4.0): you may share and adapt it, including commercially, if you credit the source. Aggregated results only; no store names. CC BY 4.0 ↗
Suggested credit: “Specoria, Agent Readiness Index 2027 (measured October 6, 2026), specoria.com/index/, CC BY 4.0”.
Method
Each store was measured with the 12 checks of the first version (v1) of Specoria’s free agent-readiness test. Scores run from 0 to 100; weights and the full list are in the appendix. The current test uses 27 checks.
- Technical readability: AI bot access (robots.txt), sitemap, llms.txt, whether the homepage answers automated requests and whether the price can be read without running JavaScript.
- Product and policy data: schema.org Product and Offer data on the product page, brand and GTIN/MPN, and shipping and return terms in machine-readable form.
- Two vantage points: The same scan ran from a home connection and from cloud egress, showing how security layers treat cloud-originated traffic too.
Scoring: each check earns its weight if it passes and half if it warns; checks that cannot be measured drop out. Weights total 100 points. A store is scored when a product page was found and at least 8 checks could be measured.
Weight by area: Access 35, Product data 45, Organization and policy 20.
Sample and its limits
English-language edition: 483 stores were listed and 471 measured (12 were dropped because the address redirected to another site or did not respond). A product page was found for 343. Sources:
- Well-known retailers and brands (named from knowledge or found among the 5,000 most visited sites in the Tranco ranking): 163
- Further storefronts found by screening a random draw of Tranco ranks 5,000-150,000: 86
- Merchants from the free Store Leads city reports for Shopify, WooCommerce, BigCommerce and Magento: 222
- The sample is not random. In the Türkiye edition, stores came from well-known brands and platform reference lists; in the English-language edition, from well-known brands, a popularity ranking filtered to online stores and platform city reports. Larger and better-known stores are therefore likely over-represented; read the figures as “share of stores in the sample”, never as a picture of every online store in a market.
- Stores that block automated access cannot be measured, so product-level results flatter the well-known brand group.
- It is one measurement on one day (October 6, 2026). Stores change; a store’s level today may differ.
- It used the 12 checks of the free test’s first version (v1). Today’s test (v2) uses 27 checks with different weights, so index levels aren’t comparable with a score from today’s test.
- Some stores block requests from cloud and data-centre networks. That is why the scan also ran from a home connection; a block seen from only one vantage point is not the same as a block for every agent.
- Specoria built the test, chose the sample and published the index. No store paid to be included.
Editions
This is the 2027 edition, measured on October 6, 2026. A new edition will be published here with its own date; past editions are not rewritten.
Full report (PDF)
The PDF has the findings in full, a fix list ranked by score gain with code examples, and the appendices. Free, no form.
Download the PDF Türkiye edition (Turkish PDF) →
Want to hear when the next edition is out? Leave your email on the report page.