What happened
Lantern launched a platform built around one question: when an AI shopping portal reads your product catalog, what does it actually understand? The system uses AI agents to predict how those portals interpret each product, flags what's constraining visibility, and applies fixes across product pages and the catalog. It rolls up to an agent-readiness score — a single number for how well your products surface in AI-driven shopping.
The feature list, per Practical Ecommerce's July 7 tools roundup: agent-ready scoring, AI visibility tracking, product-level analysis, and category benchmarking. Pricing wasn't disclosed, which for a launch aimed at merchants is worth noting on its own.
The launch matters partly as a data point about the category. Tools for making product data legible to AI buyers are multiplying fast, and categories don't grow without demand behind them. Enough merchants have discovered their catalogs are unreadable to AI shopping surfaces that diagnosing the problem is now a business model.
Why this matters
Strip the branding and "agent-readiness" is a checklist, not a mystery. An AI shopping agent reads whatever survives a crawl: attribute completeness, specs that exist as on-page text rather than trapped in images and PDFs, markup that agrees with the visible page, product names that identify the item without needing the surrounding page for context. Every one of those is checkable without a subscription, and most of them are fixable with catalog work you already know how to do.
The part that earns real skepticism is "applies fixes." Automated changes to live product data are how catalogs get quietly mangled — one bad pattern applied confidently across an entire catalog is worse than the visibility problem it was fixing, and you may not notice for weeks. Scoring is low-risk; auto-remediation is not. If you trial Lantern or anything like it, keep a human between the recommendation and the catalog.
The deeper point stands whatever you think of the tool. AI portals are already forming an opinion of your catalog, product by product, whether or not you've ever looked at what they see. The merchants treating that as an open question are behind the ones treating it as a work queue.
What to do about it
Run the manual agent-read test on your top 20 products
Paste a product URL into ChatGPT or Gemini and ask it to extract the price, the key specs, and what makes the product different from an obvious rival. Where it fails or makes things up, your page — not the AI — is usually the problem. Log every failure; that list is your agent-readiness score, free.
Pull specs out of images and PDFs
Anything that exists only in a spec-sheet image or a downloadable PDF is invisible to most AI readers. Rebuild those tables as HTML on the product page. It's tedious, and it's the single highest-yield agent-readiness fix for most catalogs we look at.
Verify markup agrees with the page
Run your best sellers through a schema validator and compare the output against what the page displays. Price, availability, or identifiers that disagree between the two teach an agent to distrust everything else you've marked up — a mismatch costs more than a gap.
Gate any auto-fix feature you adopt
If you do buy a scoring platform, export its proposed changes, review a sample by hand, and push them through your normal catalog process. A wrong score is harmless; a wrong fix applied at scale is a cleanup project.