I wrote the agent-readiness track as one section inside my Black Friday 2026 SEO prep, four bullet points buried under a bigger seasonal checklist. Then I kept running those same four checks on every affiliate page I launched after that, not just the holiday ones. So I pulled it out. This is the standalone version, the exact gate I run before a new affiliate or product page goes live on any of my sites.
It's not a ranking checklist. Classic SEO still decides whether a page shows up in a results list. This one decides whether a shopping agent, ChatGPT, Gemini, Perplexity, even Google's AI Mode, considers the page at all before a human ever lands on it. Those are two different gates now, and I treat them as separate passes.
Why a checklist beats a vibe check
Agent selection is closer to pass/fail than classic ranking is. A page with great copy and a thin Product schema doesn't rank lower, it often doesn't get considered at all. I learned this the slow way, publishing pages that read well and watching them get skipped anyway. The fix wasn't better writing, it was going back and filling in markup I'd assumed was optional.
The hard part is that nothing tells you a page failed this way. Search Console will flag a broken canonical. A schema validator will flag malformed JSON-LD. Nothing flags "this listing is technically valid but an agent still skipped it," you only find out by noticing a product never shows up when you test it directly inside ChatGPT or Gemini. That silence is exactly why I don't trust my own read of a page anymore, I trust the checklist.
So I stopped trusting my own judgment on "is this page agent-ready" and built a fixed checklist instead. Four gates, checked in order, before I ever hit publish. If a gate fails, the page waits. I covered the broader shift from classic SEO to this agent layer in SEO in the age of Super Intelligence, this post is the narrow, applied version, one checklist, run the same way every single time.
This sits downstream of my general pre-launch pass too, answer-first copy, FAQ schema, internal links into a topic cluster. That part still matters for classic rankings. The four gates below are what I added on top of it once I started treating agentic commerce as a separate audience from a human reader scanning a search results page.
The four gates, in order
I run these in a specific sequence because each one depends on the one before it. No point checking price accuracy on a page that doesn't have valid Product schema yet, there's nothing for the price field to attach to.
| Gate | What I check | How it fails silently |
|---|---|---|
| 1. Schema.org Product | Valid JSON-LD with name, image, price, availability, brand, GTIN, aggregateRating if I have reviews | Passes a basic validator but is missing a field an agent actually reads, page gets skipped with no error anywhere |
| 2. GTIN coverage | Every product has a real GTIN, UPC, EAN, or ISBN, not a blank field or a placeholder value | Schema validates fine without it, GTIN is optional per spec, but agents treat it as the primary product-matching key |
| 3. Merchant Center feed | Feed entry exists, matches the live page's price and availability, no mismatched IDs | Feed can sit stale for weeks with zero warning in the dashboard if nothing actively breaks the format |
| 4. Price accuracy | Displayed price matches the feed, matches schema, matches what the merchant actually charges at checkout | A sale badge that doesn't match tracked price history reads as noise, not a deal, to anything cross-checking it |
The JSON-LD I actually ship
This is the trimmed version of what goes on a product page before anything else happens. I keep it in a snippet file and fill in the variables per product rather than writing it from scratch each time, that's the single biggest source of missing fields I used to catch in review.
{
"@context": "https://schema.org/",
"@type": "Product",
"name": "{{product_name}}",
"image": "{{product_image_url}}",
"brand": { "@type": "Brand", "name": "{{brand_name}}" },
"gtin13": "{{gtin}}",
"offers": {
"@type": "Offer",
"url": "{{page_url}}",
"priceCurrency": "{{currency}}",
"price": "{{price}}",
"availability": "https://schema.org/{{availability}}",
"priceValidUntil": "{{price_valid_until}}"
}
}
The two fields I see missing most often on other sites: gtin13 and
priceValidUntil. Neither is required for the markup to validate. Both are exactly what an
agent cross-checks before deciding the listing is trustworthy enough to surface.
GTIN: where launches quietly break
GTIN coverage is the gate I fail most often on a first pass, almost always on affiliate pages for smaller or regional brands that never had a proper barcode assigned, or where the merchant's own product feed left it blank. Schema.org doesn't require it. Agents still use it as the main key for matching a listing across retailers, so a missing GTIN doesn't make the page invalid, it makes the page invisible to anything that's trying to confirm the product is real before recommending it.
My fix isn't elegant. For products without a manufacturer GTIN, I pull it from the merchant's own product feed if they publish one, or I flag the product for manual lookup rather than publishing with a blank or guessed value. A wrong GTIN is worse than a missing one, it tells an agent two different products are the same item.
I keep the flagged list in a simple spreadsheet, one row per product, GTIN status, where I looked, who I'm waiting on. It's not automated yet. Most weeks it's two or three products, small enough that a spreadsheet beats building tooling for it, and I'd rather see the gap plainly than hide it behind a dashboard I stop checking.
I once copied a GTIN from a near-identical variant product to get a page out the door faster, meaning to fix it later. I didn't fix it for three weeks. During that window the page was matched against the wrong product in a price-comparison surface, showing a price that wasn't mine next to my listing. Now a blank GTIN blocks publish entirely in my checklist, a wrong one is worse than blank.
Merchant Center and price accuracy: the part that doesn't end at launch
The first three gates are mostly one-time work per page. This last one isn't. A feed that matched the page perfectly on launch day can drift within a week if a merchant changes price and I haven't re-synced the feed, or if stock runs out and the feed still shows it available.
So gate four isn't just a launch check, it's a recurring one. I run a lighter version of the same price and availability comparison weekly on everything already live, not just new pages. That's the same habit I described scaling up for the holiday peak in the Black Friday piece, checked daily during a four-day sale window instead of weekly. The checklist doesn't change between a normal Tuesday and Black Friday, only the frequency does.
| Mismatch | What I check it against | How often |
|---|---|---|
| Displayed price vs. feed price | Google Merchant Center diagnostics panel | Weekly, daily during a sale window |
| Displayed price vs. actual checkout price | Manual spot-check, I buy or add-to-cart test a sample myself | At launch, then monthly |
| "Sale" badge vs. tracked price history | My own price-history log, not the merchant's claimed discount | At launch, then whenever the badge changes |
| Availability flag vs. real stock | Merchant Center feed status | Weekly, daily during a sale window |
That price-history check matters more than it sounds like it should. An agent that cross-references a claimed discount against tracked history and finds the "sale" price is the same as last month's regular price doesn't just ignore the badge, it treats the whole listing as less trustworthy going forward. I'd rather show no discount badge than a false one.
I keep my own price-history log separate from the merchant's feed on purpose. If I only trusted the feed's claimed "was" price, I'd be checking whether the merchant's marketing matches their own marketing, not whether the discount is real. Tracking price independently, even with a simple scraped log updated daily, is the only way I've found to catch a badge that's technically accurate per the feed but misleading in practice.
The checkout spot-check is the one step in this whole gate I haven't managed to automate, and honestly don't want to. Clicking through to an actual cart once a month on a sample of products catches things a feed comparison never will, a currency mismatch on a regional variant, a shipping surcharge that changes the effective price, a merchant quietly discontinuing a size or color the feed still lists as available. Thirty minutes once a month is cheap insurance against publishing a page an agent recommends and a shopper then can't actually buy.
Can I open the live page, the schema validator output, and the Merchant Center feed entry side by side and confirm all three show the exact same price and availability, right now, not from memory? If any of the three disagree, the page doesn't go live yet.
Key takeaways
- Agent selection is closer to pass/fail than ranking is, a page can read well and still get skipped entirely over one missing schema field.
- The four gates run in order: valid Schema.org Product markup, GTIN coverage, a Merchant Center feed that matches the live page, and price accuracy checked against real history.
- GTIN is optional per the Schema.org spec and treated as a primary matching key by agents anyway, a wrong GTIN is worse than a missing one.
- Price accuracy isn't a launch-day check, it's a recurring one, feeds drift within days of a merchant changing price or stock.
- I'd rather delay a publish than ship a page where the displayed price, the schema, and the feed don't all agree.
What I'll do next
I'm turning the side-by-side price check into a scheduled routine, the same pattern I described in automating my SEO checklist with Claude routines, so the weekly comparison runs on its own instead of me opening three tabs manually. I'm also building a GTIN-coverage report across every live product page, so I can see the gap before a gate fails on a specific launch instead of catching it one page at a time.
FAQ
Do I need a GTIN for every product, even small or regional brands? +
How is this different from a regular pre-launch SEO checklist?
How often should I re-check the Merchant Center feed after launch?
What happens if I publish with an incomplete Product schema?
Is this checklist only for Black Friday or seasonal sales?
Want this built for your team?
I design AI agents and growth automation that run without babysitting. If that sounds useful, let's talk.
Get in touch โ