Black Friday isn’t a discount event anymore. It’s a prompt-volume event. Between November 22 and 29, AI shopping prompts rise +340% above the October baseline, retrievers lock their citation decisions early, and every Shopify merchant who edits schema inside the 48-hour window before BF Tuesday watches their citation share drop for the exact 72 hours they wanted it to peak. This is the sequencing playbook: what to ship by when, what to freeze, and how to monitor without panicking.
What actually changed between BFCM 2023 and 2025
In 2023, the Black Friday marketing calendar was built around three channels: email, paid social, and (distantly-third) organic search. Prompt-based shopping was a rounding error. By 2025, prompt volume on Perplexity, AI Mode, ChatGPT, and Claude hit parity with organic search during BF week for categories like beauty, furniture, and supplements. Shoppers opening the week by asking “best category BF deal under $price” on Claude or Perplexity became normal enough that retrievers started pre-caching category answers in the week before the peak.
That pre-caching is the detail that rewrites the merchandising calendar. If retrievers are locking their answer candidates during November 22-24 and serving them for the next five days, every schema tweak, FAQ rewrite, or URL change after the lock window is a net-negative action — you’re invalidating a cache that was already about to cite you and forcing the retriever to re-crawl under time pressure it doesn’t have.
The four phases — and what each one costs if you get it wrong
Phase 01 — Baseline (November 1-14)
This is the only phase where schema work is cheap. Crawlers re-fetch normal cadence, retrievers aren’t under prompt-volume pressure, and any FAQ, Product, Offer, or Review schema fix you ship now has ~14 days to propagate, validate, and earn its citation slot before the retrieval lock. Ship your Q4 schema changes here or not at all. We audit Shopify merchants every November and the ones who shipped big schema refactors in early November sat 8-14 percentage points above the merchants who pushed the same refactors in mid-to-late November.
Phase 02 — Warm-up (November 15-21)
Gift-oriented prompts start. “Best gift forpersona under $price,” “gift ideas for relationship who loveshobby,” and early deal-finder prompts (“early BF deals on category”) begin showing up in AI-engine samples by November 15-16. This is the corroboration window. Retrievers are pulling citation candidates and cross-checking them against third-party context — Reddit threads, YouTube video descriptions, reputable buying guides. If your brand name isn’t in the third-party context, you get cut. Push one or two genuine Reddit contributions in the right subreddits, coordinate with creators on video corroboration, and publish one comparison guide that interlinks with your top PDPs. The window closes around November 21.
Phase 03 — Peak (November 22-29)
The retrieval-lock window. This is where discipline beats cleverness. Engineers want to ship a “small fix.” Merchandising wants to swap the FAQ copy to push a specific SKU. Content wants to rename a landing page. Every single one of those reflexes is wrong. Retrievers that had you in the candidate pool on November 22 will hold that candidate through cache invalidation events during BF week; if you change the page, the cache drops and under prompt-volume pressure the retriever picks the next candidate (a competitor) rather than waiting for the re-crawl.
The peak phase is monitoring-only. Sample the top-30 BF prompts across all four engines once on November 22, once on November 25 (after BF Monday), and once on November 29. Log citation share and rank per engine. If you see a sudden drop that isn’t tied to a known technical issue, escalate to the GEO lead — not to the content or dev teams. The fix is almost always “stop shipping things,” not “ship something.”
Phase 04 — Cool-down (November 30-December 1)
Post-purchase prompts spike. “Return policy forbrand,” “how long does brandtake to ship,” “can I cancel my BF order.” If your MerchantReturnPolicy schema and public returns page render cleanly, you pick up a second citation wave. If they’re broken, retrievers answer the question with competitor policy pages or with “we couldn’t find a clear answer,” which damages brand trust exactly at the moment the shopper is deciding whether to keep or return. See our return-rate signal piece for the mechanics.
The schema freeze list
When we say “freeze,” we mean these specific fields are untouchable from November 22 onward. Any edit to any item on this list after the lock window opens requires GEO-lead sign-off, not just a dev or merchandiser ticket.
- Product schema: name, brand, sku, gtin, offers.price, offers.priceCurrency, offers.availability
- AggregateRating and Review schema on every PDP with ≥50 reviews
- FAQPage schema on PDPs, collection pages, and /pages/faq
- MerchantReturnPolicy schema and /pages/return-policy content
- Canonical URLs — no redirect changes, no slug edits
- Primary H1 and title tag on every page in the BF campaign
- llms.txt and ai-sitemap.xml (regen once at T-72 and leave alone)
What to measure during the lock window
Three metrics, sampled three times across BF week, logged per engine. Everything else is vanity.
Citation presence — is your brand cited at all for each of your top-30 BF prompts? This is a binary per prompt per engine. If presence drops on a prompt where you were present 24 hours earlier, investigate: is the page still reachable? Has schema validation broken? Has a competitor corroboration wave moved you?
Citation rank — when you’re cited, are you in the top 3 answer sources or buried below the fold? Rank drops inside the lock window almost always come from competitor content gaining corroboration, not from your pages losing quality. The response is to push one more corroboration piece, not to touch the PDP.
Engine mix — which engines are citing you, and is the mix stable? If Perplexity drops you but ChatGPT and Claude hold steady, that’s a Perplexity-specific ranking nudge (usually corroboration weight) rather than a broad retrieval problem. If you drop across all four engines simultaneously, it’s almost certainly a technical problem on your site: 5xx, JSON-LD parse error, or a middleware misfire stripping schema.
Post-BF recovery posture
On December 1, unfreeze. Ship the backlog of small changes that were held during the lock window. Run the first full-site schema re-validation since November 22. Publish any post-BF case-study content that references the week (this is the one window where BF-themed content actually performs well in retrieval — shoppers searching for “how did brand do during BF 2026”). Schedule the next big schema or content push for January — December is a second elevated prompt-volume period (holiday gifting, end-of-year gifting) and touching schema inside that window risks the same cache invalidation on a smaller scale.
The bottom line
Black Friday citation share is won or lost before the peak starts. Ship schema in early November, corroborate mid-month, freeze at T-48, monitor through the peak, and unfreeze on December 1. Teams who treat BF like a live-ops event with constant edits lose citation share; teams who treat it like a locked-cache window gain share because they’re stable while competitors thrash. The discipline is cheap, the upside is 4-12 points of citation share, and the only way to lose the whole thing is by shipping one “small” fix inside the lock window.