Black Friday in the Agent Era: What to Fix Before November

Black Friday lands on November 27 this year. You have four months, and almost everything worth doing in them is unglamorous data work that has to be finished before the traffic arrives.

Here's why it's worth your September.

Start with the numbers, and start skeptically. Adobe reported that traffic to US retail sites from generative AI tools rose 693.4% over the 2025 holiday season year over year, and 670% on Cyber Monday alone. Adobe also added the caveat that gets left out of most write-ups: the base of users remains modest. A 693% increase off a small base is still a small number, and anyone quoting that figure without the caveat is selling something.

What makes it harder to wave away is the second set of numbers, which are about orders rather than clicks. On Shopify's Q1 2026 earnings call, president Harley Finkelstein said AI-driven traffic to Shopify stores grew eightfold year over year, with orders from AI searches up nearly 13x and new buyers ordering through AI channels at close to twice the rate of other channels. Adobe's Q1 2026 traffic report, covered by Search Engine Journal, put AI-referred retail traffic up 393% year over year, with revenue per visit 37% above non-AI traffic and conversion running 42% better in March. Twelve months earlier that channel was the worst-performing one Adobe tracked.

One more, with an asterisk. Salesforce put global Cyber Week 2025 spend at $336.6 billion, up 7%, with AI and agents influencing $67 billion of it and roughly 20% of all orders. Its headline stat, that retailers running Agentforce and their own branded agents grew sales 32% faster than those without, is a vendor measuring its own product. Worth knowing, not worth planning around. It's also a different thing from what this post is about: Salesforce is mostly counting chat and service agents on the retailer's own site. An assistant that recommends you to a stranger who has never heard of you is a separate mechanism with separate requirements.

Deal hunting is the ideal assistant query

Think about what makes a query worth delegating. It's tedious, it's comparative, and it has a checkable answer. "Find me a merino base layer under $120 in medium that ships before the 22nd" is five constraints across an unknown number of stores, which is exactly the kind of work nobody enjoys doing in twelve browser tabs.

Google's AI features documentation describes what happens next: a "query fan-out" technique, "issuing multiple related searches across subtopics and data sources" before assembling a response. Your hero image, your countdown timer, your carefully staged lifestyle photography, none of it participates. What participates is your data.

What actually gets checked

When something compares your offer against three others, it's looking for a short list of facts:

• Price, and specifically whether the sale price exists in your data rather than only in a theme rendering.
• Availability, per variant, not per product.
• Shipping cost and delivery speed. Google's product structured data guidance calls out shipping details, especially free shipping, so shoppers understand total cost.
• Return policy, including fees and timeframe.
• Ratings and review counts.
• Whether all of the above matches what's visible on the page. Google's rule is that structured data has to reflect the visible content, and a price in your markup that disagrees with the price on screen is a problem you want to find in September rather than on the 27th.

Shopify Catalog already exposes most of this on your behalf. Per Shopify's own explainer, it standardizes products into a universal taxonomy, verifies pricing and inventory in real time, and syndicates to ChatGPT, Copilot, Google AI Mode and the Gemini app, with merchants included by default. Which is the uncomfortable part. Your catalog is being read whether or not you've thought about it, and the version being read is whatever your data says today.

The failure modes that are specific to BFCM

Discount codes are the big one. If your 30% off applies at checkout or requires a code, the price a machine reads is the full price. Shopper needs under $150, your data says $180, you're out of the comparison before your creative is ever seen. Sale prices belong in the product data, not only in a banner.

Sold-out variants marked available. On a normal Tuesday this is a minor annoyance. During Cyber Week, when a size sells through in ninety minutes, stale availability data means you get recommended for something you can't ship, and the shopper's next question to the assistant is "find me another one."

Shipping cutoffs that live in an image. "Order by Dec 18 for Christmas delivery" as text on a homepage banner is invisible to everything that isn't a human with working eyes. Put it in text, on the shipping policy page, in a sentence.

A deal page created on November 20. This one predates AI entirely and it's still the most common mistake, which is why we wrote the BFCM SEO checklist around a permanent URL that stays live all year. A page with no crawl history has no chance in the most competitive four weeks of the year.

And the sneaky one: rewriting product descriptions for holiday urgency and stripping out the specs in the process. "Our biggest deal of the year" replacing "100% merino, 190gsm, machine washable" is a straight downgrade for anything trying to match your product against a five-constraint query.

The timeline

September is data cleanup, and it's the month that matters. Run a full site audit. Fix broken links, missing or duplicate metadata, pages with no structured data, and variant-level availability. Set your tracked query baseline now, while traffic is normal, because a baseline taken in November tells you nothing.

October is content and metadata. Seasonal titles and descriptions in bulk across products and collections. FAQ coverage on the pages that answer shipping, returns and sizing, because those are the questions an assistant checks before recommending anyone. Structured data across the catalog. Deal hub live, linked from the homepage and your best-performing product pages, populated with placeholder offers if the real ones aren't finalized.

November is a freeze. Change prices and inventory. Do not change templates, URL structures or theme code in the four weeks that pay for your year. Watch your tracked queries a couple of times a week and fix factual errors if an assistant is describing your policy wrong.

December is the post-mortem, and the one rule is don't delete anything. Keep the deal page, repoint it, and compare your tracked query set before, during and after.

Track the questions, not just the rankings

Search Console will not tell you what ChatGPT said about you. Nothing will, unless you go and look.

Pick 15 to 25 questions in your category, phrased the way a shopper would actually type them, with the constraints they'd actually apply. Not "best running socks" but "black friday deals on merino running socks under $30." Check them in September for a baseline, weekly through November, once in December. Seokai's AI Visibility tracker runs that on a schedule so you're not doing it by hand at 11pm on the 27th.

The number that matters isn't a position. It's whether you get named at all, and whether what's said about you is true. A confidently wrong return policy in an AI answer costs you more than a missing citation.

Your deal page is for humans. Your product data decides whether a human ever sees the deal page. That's the whole shift, and it's the reason September is the month, not November.

You can see where your store currently stands on the machine-readable side with the audit and AI Readiness check in Seokai.