Description
AgenticTrack scores how ready your WooCommerce products are for AI shopping agents like ChatGPT, Gemini, Claude, and Perplexity — then helps you fix the gaps, get your store discovered, and track the orders agents bring. The audit, the core fixes, and the discovery tools are all free; AI fixes and agent-order analytics are Pro.
AI shopping agents are starting to read, compare, and buy from stores directly through structured surfaces like the WooCommerce MCP integration, the WordPress Abilities API, and the emerging agentic commerce protocols (ACP, UCP, MCP). These agents don’t read your storefront design or marketing copy — they read your structured product data. Products with thin descriptions, missing identifiers, or incomplete structured data are invisible — or unsellable — to those agents.
The audit is free and read-only. It scans your published products and gives each one an agent-readiness score from 0 to 100, so you can see exactly which products AI agents can understand and which are holding you back. The audit never changes your data and makes no external API calls.
What it checks for every product:
- Product identifiers — SKU and GTIN/UPC/EAN (the native WooCommerce global unique ID, plus common third-party fields) that AI agents need to match your products.
- Structured attributes — material, size, color, weight, dimensions, and other machine-readable facts.
- Description quality — flags thin copy and marketing-heavy text that lacks the factual specifications AI agents rely on.
- Price & inventory — a valid price and an unambiguous stock status.
- Structured-data completeness — whether each product has the fields needed for rich, valid Product JSON-LD schema.
You get a per-product agent-readiness score, an overall store score, a sortable product list with the exact issues found, and a summary of the top issues across your store.
Fix the gaps (free)
The Fixer repairs the issues the audit finds, writing to your own store through the WooCommerce API — no catalog is routed through a middleman. The manual fixes are free; only the AI-powered rewrites are Pro. Every change is built to be safe:
- Preview first — you see exactly what will change, before anything is written.
- Approve before apply — nothing is saved until you click Apply. Apply one fix at a time, or use “Apply all safe fixes” with a live progress readout.
- One-click undo — the previous value is snapshotted before each write, so a single click restores it.
Free fixes — all on the same safe rails, no AI and no external calls:
- Structured data — fills the short description (schema summary) from your existing long description; enter a missing price or assign a category right in the panel. It never fabricates data.
- Description editor — write or expand a thin product description in-place, saved exactly as you type.
- Attribute normalization — tidies the naming, casing, and spacing of your existing custom attributes so agents read consistent facts.
- Identifier setup — lets you enter a real GTIN/UPC/EAN and SKU, saved to the correct WooCommerce field. Identifiers are never invented — you provide the value.
AI fixes (Pro) need no API key of your own — with Pro active they run through the AgenticTrack AI service, strictly capped per run. You generate a rewrite, review (and edit) it, then save. Prefer to use your own AI account? Pick a provider in Settings — OpenAI (ChatGPT), Anthropic (Claude), Google Gemini, or OpenRouter — add that provider’s key, and requests go straight to it:
- Rewrite description (AI) — rewrites thin or marketing-heavy descriptions into factual, spec-rich copy that agents can read. You approve each rewrite before it is saved.
- Suggest attributes (AI) — proposes machine-readable attributes (material, color, size, …) from the product’s own title and description. You approve before they are added.
Make your store discoverable to agents (free)
Agent Access is the site-level companion to the per-product audit, and it is free:
- AI-crawler rules — allow the major AI and shopping-agent crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and more) in your robots.txt with one toggle. Nothing is written to disk — it uses the standard WordPress robots.txt filter.
- Product-aware llms.txt — publishes an
/llms.txtmap of your catalog (categories and every product with price, SKU and a factual line) so agents crawling it can discover what you sell. Previews are shown right inside the plugin.
See which orders came from AI agents (Pro)
Agent Orders attributes the sales that bypass normal analytics — orders placed by AI shopping agents through headless and API checkouts — so you can measure agentic revenue:
- Two tiers, shown separately: Verified — the order came through a UCP/ACP agent checkout that presented a signed AP2 payment mandate (cryptographically provable, so it carries no hedging); and Inferred — a best-effort estimate from checkout signals (agent user-agents, headless/API checkouts, referrers) with a confidence level.
- Agent orders and revenue over time, a by-agent breakdown, and the share of your orders coming from agents.
- Only the Inferred tier is an estimate, and it is labelled as such. Verified attribution is set only when an agent checkout hands over a signed payment mandate — heuristics can never promote an order to Verified. Most stores have no agent-checkout traffic yet, so Verified showing 0 is normal and expected. Detection runs from install, so the data is ready the moment you go Pro.
Why AgenticTrack
- Your data, your store — the audit and the free fixes run entirely on your own store through the WooCommerce API and make zero external calls. AI fixes call out only when you click Apply, and only send that product’s title and description (see “External services”).
- Built for the agentic commerce shift — checks and repairs the exact product-data signals that determine whether AI shopping agents can find, understand, and recommend your products.
- Fast on large catalogs — batched, paginated scanning that stays fast and reliable even on stores with thousands of products.
- Developer-friendly — the audit checks and the fixes are both extendable with your own rules through standard WordPress filters.
- HPOS compatible.
Get your WooCommerce store’s agent-readiness score free, and start closing the gaps before AI agents become a meaningful share of your orders.
External services
The free audit makes no external calls — it runs entirely on your own store. Nothing is sent off-site unless you use a Pro feature that requires it, and only when you opt in.
-
License validation. When you save an AgenticTrack Pro license key, the plugin sends the key and your site URL to the AgenticTrack service at agentictrack.ai to confirm the license, then re-checks about twice a day. No product or customer data is sent. This happens only if you use a Pro license. See the AgenticTrack privacy policy and terms at https://www.agentictrack.ai.
-
AI fixes. When you click Apply (or Generate) on an AI description rewrite or AI attribute suggestion, the product’s title and description are sent to an AI service so it can draft the result. This happens only on that click; no data is sent for the free fixes or the audit. Which service depends on your choice under Settings AI provider:
- AgenticTrack AI (default, no key needed): the request goes to the AgenticTrack AI service at agentictrack.ai; your site URL and license key are included so the service can authorize it. See the AgenticTrack privacy policy at https://www.agentictrack.ai.
- Your own provider key: if you choose a provider and paste its key, requests go directly to that provider and never touch the AgenticTrack service. Supported providers and their policies: OpenAI — https://openai.com/policies/ ; Anthropic (Claude) — https://www.anthropic.com/legal/privacy ; Google Gemini — https://ai.google.dev/gemini-api/terms ; OpenRouter — https://openrouter.ai/terms and https://openrouter.ai/privacy.
-
Agent Orders dashboard sync (Pro). With an active Pro license, when an order is detected as placed by an AI agent, the plugin reports that detection to the AgenticTrack service at agentictrack.ai so it also appears in your AgenticTrack account dashboard. What is sent: the order number, order total and currency, your site URL, your license key, and the detection evidence (which signals matched and the tier) — no customer names, emails, addresses, or payment details, ever. Human orders are never reported. The call is non-blocking and can never affect checkout. Disable it (keeping detection purely local) with the
arwc_agent_order_sync_enabledfilter. See the AgenticTrack privacy policy at https://www.agentictrack.ai/privacy.
Screenshots
FAQ
Does this change my products?
The audit never does — it is completely read-only. The Pro Fixer (and the free structured-data, attribute, and identifier fixes) can change products, but only when you explicitly click Apply. Every change is previewed first, and one-click undo restores the previous value.
Can it invent a GTIN or barcode for me?
No. Identifiers are real-world facts, so the plugin never generates them. When a GTIN/UPC/EAN or SKU is missing, it gives you a field to enter the real value, validates it, and saves it to the correct WooCommerce field.
What is free and what is Pro?
Free: the full audit and store score, all manual fixes (structured data, price, category, the description editor, and GTIN/SKU), the Agent Access module (AI-crawler robots.txt rules and a product-aware llms.txt), live re-scoring, a 7-day view of AI Sales, and a few AI fixes per day. Pro: unlimited AI fixes (description rewrite and attribute suggestions), the full 90-day AI Sales history with CSV export, and store-wide bulk fixing with no product cap. The manual/free features make no external calls; AI fixes call the AI service only when you click Generate/Apply.
Do I need to pay or add an API key to use the fixes?
The manual fixes — structured data, price, category, the description editor, attribute normalization, and identifier setup — are free and make no external calls. The AI description rewrite and attribute suggestions are a Pro feature — but with Pro active they need no API key of your own, running through the AgenticTrack AI service. If you’d rather use your own AI account, choose a provider under Settings — OpenAI (ChatGPT), Anthropic (Claude), Google Gemini, or OpenRouter — add that provider’s key, and requests will go directly to it instead.
How accurate is Agent Orders attribution?
It depends on the tier, and the two are shown separately. Verified orders came through an agent checkout (UCP/ACP) that presented a signed AP2 payment mandate — that is cryptographically provable, so those are not estimates. Inferred orders are a best-effort guess from checkout signals — the agent’s user-agent, whether the order came through an API/headless checkout rather than a browser, and referrer/attribution data — each with a confidence level; this tier will get sharper as agentic-commerce standards become common. Most stores have no agent-checkout traffic today, so the Verified count will simply be 0 — that is correct, not a bug. Detection runs on your own store; with an active Pro license, detected agent orders (order number, total, currency — never customer data) are also reported to your AgenticTrack account dashboard (see “External services”).
Does it work with large catalogs?
Yes. Scanning is batched and paginated, so it stays fast and reliable even on stores with thousands of products.
What are AI shopping agents, and why does agent-readiness matter?
AI shopping agents are assistants like ChatGPT, Gemini, Claude, and Perplexity that browse, compare, and increasingly buy products on a shopper’s behalf. They read structured product data, not your storefront design. If your product data is incomplete, agents may skip your products — so agent-readiness directly affects whether your store shows up in agentic commerce.
How is the structured data check measured?
WooCommerce and most SEO plugins already emit Product JSON-LD automatically, so this check scores whether each product has the fields that make that schema rich and valid (image, price, description, identifier, category) — measured from product data, with no page fetching.
What data leaves my site?
The audit and the free fixes send nothing off-site. If you enter a license key, that key and your site URL are sent to agentictrack.ai to validate it. When you click Apply on an AI fix, the product’s title and description are sent to the AgenticTrack AI service (or directly to the AI provider you configured with your own key). With an active Pro license, orders detected as placed by AI agents are reported to your AgenticTrack dashboard — order number, total, and currency only, never customer names, emails, addresses, or payment details. See the “External services” section above.
Reviews
There are no reviews for this plugin.
Contributors & Developers
“AgenticTrack – Agentic Commerce Readiness for WooCommerce” is open source software. The following people have contributed to this plugin.
Contributors
Translate “AgenticTrack – Agentic Commerce Readiness for WooCommerce” into your language.
Interested in development?
Browse the code, check out the SVN repository, or subscribe to the development log by RSS.
Changelog
1.3.0
- New: bulk-add SKU & GTIN from a spreadsheet — download a CSV of all your products, fill in the SKU/GTIN columns, and upload it to set them across your whole catalog at once. Every change is validated (8–14-digit GTINs, unique SKUs) and undoable per product. The fastest path from a mid-range score to 85+ on large stores.
- Improved: clearer positioning of the MCP / Abilities endpoint as the machine-readable API layer of agent readiness (alongside robots.txt and llms.txt) — forward-looking infrastructure for AI agent/app integrations.
- New: “Agent preview” on the Fix page — see the exact structured data an AI shopping agent receives for a product, with missing fields highlighted, so you understand at a glance why each fix matters.
- New: inline quick fixes in the product editor — set a featured image or enter GTIN/SKU right in the Agent Readiness sidebar box, without opening the Fixer; the score updates on the spot.
- New: a clear licence status line under the key field (Pro · Active / validation failed), so you always know whether your key worked.
- New: a “What does Good (80+) mean?” explainer on the Audit page.
- Improved: the “leave a review” prompt now retires itself after two weeks.
- New: free stores now get a few AI fixes every day (powered by gpt-4o-mini) — a daily taste of the AI description rewrite and attribute suggestions; Pro stays unlimited. The Fix panel shows how many free AI fixes you have left today.
- Improved: clearer AI errors — if you’re on a local/development site (where the built-in AI service isn’t reachable), the plugin now says so and points you to add your own provider key; rate-limit messages include a “try again in about N” timer when known.
- Improved: the Audit scorecard now shows a “you’re X points from Good, here are the biggest gaps” nudge so you know exactly what to fix next.
- Fixed: the “AI Sales” page is now reachable at both its canonical URL and a friendly /arwc-ai-sales alias (no more permission error from direct links).
- New: AI Sales (formerly Agent Orders) is now visible on the free plan — free stores see the last 7 days of AI-agent order attribution, so you can see agentic sales as they start; Pro keeps the full 90-day history plus CSV export.
- New: Score-history trend on the Audit page — a sparkline of your store score over time (30 days on free, a full year on Pro) so you can see progress at a glance.
- New: “Quick wins” banner on the Fix page — after an audit it highlights how many products have fast, free fixes and links you straight to the first one.
- New: set a product’s featured image without leaving the Fix page — the image fix now opens the WordPress media picker inline.
- New: the store-wide “Apply safe fixes to all products” covers the first 5 products on free (individual per-product fixes stay unlimited); Pro removes the cap.
- Improved: the product edit-screen panel now shows the same CRITICAL / WARNING / INFO severity tags as the Audit page.
- Improved: the “Get Started” guide is retired from the menu once you’ve run your first audit, to reduce clutter (still reachable to replay the wizard).
- Improved: the built-in AI service now uses a fast, low-cost gpt-4o-mini model — far more reliable than the rate-limited free models it replaces — and bring-your-own-key providers (OpenAI, Anthropic, Gemini) now default to their latest models.
- Improved: the product edit-screen “Agent Readiness” panel is now pinned to the top of the product sidebar, so it stays visible even if you’ve previously rearranged your meta boxes.
- Improved: the store-score chip briefly highlights when it refreshes after an audit, so you can see it updated even when the number is unchanged.
- Improved: the admin-bar score badge now shows the current product’s score while you’re editing a product (with a link to its fixes), instead of the store average — so it reflects what you’re looking at.
- Improved: when the AI service reports how long to wait, the rate-limit message now includes a concrete “try again in about N seconds/minutes” timer.
- Improved: clearer guidance when the free shared AI is rate-limited — instead of a bare “try again”, it now explains the shared free models are busy and points you to add your own AI provider key (OpenAI, Anthropic, Gemini, or OpenRouter) for dedicated limits.
- Improved: deleting a static llms.txt now verifies the live /llms.txt afterwards — if another plugin or a server rewrite rule is still serving it, you’re told clearly instead of getting a false “all good”. The Agent Access page also shows whether your live /llms.txt is genuinely AgenticTrack’s map.
- Improved: your store score in the header updates live after “Apply safe fixes to all products”, not just after a manual re-audit.
- Improved: the Agent Orders page now has a plain-language “Verified vs Inferred” explainer for merchants new to agent-checkout standards (ACP/UCP/AP2).
- Improved: the “leave a review” prompt now appears only on the Audit dashboard, not on every screen.
- New: Fix your whole catalog at once — an “Apply safe fixes to all products” button on the Fix page applies the free automatic fixes (schema summary + attribute tidy-up) across every published product, with a live progress readout. Each change is still saved per product and can be undone individually; AI and manual fixes are never bulk-applied.
- New: Agent-readiness score column on the WooCommerce Products list — see each product’s score (and jump straight to its fixes) without leaving your normal product workflow.
- New: One-click removal of a static llms.txt — when another plugin has left a static llms.txt on disk that shadows AgenticTrack’s product-aware map, you can now delete it from the Agent Access page instead of using FTP.
- Fixed: the AI description rewrite could show the model’s internal “thinking” instead of the finished description (and AI attribute suggestions could come back empty). The default AI model is now a non-reasoning model, and the plugin never surfaces a model’s chain-of-thought — so you always get clean, factual copy to review.
- Fixed: your store score in the branded header and admin-bar badge now updates immediately when an audit finishes, instead of showing the previous score until you navigate away.
- Improved: the Agent Access page now shows the full REST URL for detected MCP/Abilities endpoints and explains they’re API endpoints (a direct request returning JSON or a 401 is expected, not an error).
- New: Guided first-run setup wizard — on first activation, a short, on-brand, skippable wizard welcomes you, runs your first audit, shows your store’s agent-readiness score, and points you to the top issues to fix. It only ever appears once.
- New: Get Started guide — a persistent, friendly onboarding checklist (run audit review scores fix issues make your store discoverable optional Pro AI fixes) so you always know the next step. Replay the wizard anytime from here.
- New: Product edit-screen panel — each product’s agent-readiness score and its top issues now appear right in the product editor sidebar (near the top), with a one-click link into the Fixer and a clickable “see the full list” link. The score updates whenever you save the product, and saving now also keeps your store average up to date.
- Improved: Clearer, less intimidating audit screen — a friendlier empty state with a single obvious “Run your first audit” action, a safety reassurance, and plain-language guidance for new users.
- New: Agent Orders now sync to your AgenticTrack account dashboard (Pro) — orders detected as placed by AI shopping agents are reported to agentictrack.ai so you can see them alongside your scans, on any device. Only the order number, total, currency and detection evidence are sent — never customer names, emails, addresses, or payment details. Human orders are never reported. The call is non-blocking, so checkout speed and reliability are completely unaffected; disable it with the
arwc_agent_order_sync_enabledfilter to keep detection purely local.
1.2.1
- New: Bring your own AI provider — as well as the built-in AgenticTrack AI service, you can now run the Pro AI fixes through your own OpenAI (ChatGPT), Anthropic (Claude), Google Gemini, or OpenRouter account. Pick a provider under Settings, paste that provider’s key, and requests go straight to it. Usage stays capped per run.
- Improved: clearer Settings page — an AI-provider selector with a dedicated key field for each provider, and per-provider default models.
- Improved: refreshed plugin description and store assets.
1.2.0
- New: Agent Orders (Pro) — attribution of orders placed by AI shopping agents (which bypass normal analytics), split into two tiers shown separately: Verified (the order presented a signed payment mandate through a UCP/ACP + AP2 agent checkout — provable, not a guess) and Inferred (a best-effort estimate from checkout signals such as agent user-agents, headless/API checkouts and referrers). The dashboard shows agent orders, revenue, order share, a by-agent breakdown, and a 30-day trend. Most stores have no agent-checkout traffic yet, so Verified showing 0 is normal. Detection runs on every install so the data is ready the moment you upgrade.
- New: Agent Access (free) — allow the major AI/shopping-agent crawlers in robots.txt with one toggle, and publish a product-aware /llms.txt map of your catalog, with in-plugin previews and detection of conflicting plugins.
- New: Generous free tier — the full audit and score, all manual fixes (structured data, price, category, description editor, GTIN/SKU), llms.txt, and AI-crawler rules are now free. Pro is AI fixes and Agent Orders analytics.
- New: Rebuilt admin as one cohesive app — branded header with plan and store-score badges, an in-app sidebar (Audit / Fix / Agent Access / Agent Orders / Settings), grade filters on the product table, and toggle switches.
- New: AI fixes are now two-phase — generate a proposal, review and edit it, then save; nothing is written until you approve.
- Improved: fixes re-score the product live (score and issue count update in place); the previously “flagged” price, category and description items are now fixable directly in the panel.
- HPOS-safe throughout: order tracking reads and writes order meta via WooCommerce CRUD only.
1.1.0
- New: Pro license activation — paste your AgenticTrack key under Agent Readiness Settings; it is validated against agentictrack.ai and re-checked periodically, with graceful handling when a key has reached its activation limit.
- New: The Fixer — a safe, reversible write-side view alongside the audit. Preview first, approve before apply, and one-click undo for every change.
- New: Free fixes — structured-data completion (short description from your existing copy; flags missing image/price/category), custom-attribute normalization, and merchant-entered GTIN/SKU setup (identifiers are never invented).
- New: AI fixes (Pro) — factual description rewrites and structured-attribute suggestions, each previewed and approved, with hard caps on calls and tokens per run. With Pro active they need no API key of your own (provided through the AgenticTrack AI service on free models), or use your own OpenRouter key.
- New: Live OpenRouter model picker and a no-cost “Test AI key” button on the Settings page; agent-readiness score badge in the admin toolbar.
- New: “Apply all safe fixes” with a live progress readout for bulk applying the no-AI fixes on a product.
- The free audit remains read-only and makes no external API calls.
1.0.0
- Initial release: read-only agent-readiness audit and scoring dashboard.