How AI Is
Talking About Whatnot
Whatnot is the name AI says first the moment a conversation enters the live-commerce frame. The room to grow sits one step earlier, in the urgency-shaped questions buyers and sellers actually type, where the live-auction answer rarely surfaces and competitor proof carries the comparison.
What the AI conversation says today
Ask ChatGPT, Claude, or Gemini about live shopping, and Whatnot is the first name AI says.
That is a real moat. Most brands would trade a lot to own a category word inside an AI conversation. The brand work, the Seller Academy, the help-center investment, the community story: it is all landing inside the model. Inside the live-commerce frame, you own the conversation. That is the floor we are building from.
The next leg of growth sits just upstream of that win. Almost nobody types “live shopping” into ChatGPT. They type the pain.
Buyers say: “boosters are sold out everywhere”, “I missed the SNKRS draw”, “tournament Saturday, need singles fast”.
Sellers say: “fees ate my margin”, “my side hustle stopped scaling”, “card-show inventory pile-up”.
In those moments (the ones that happen before anyone knows to ask about live commerce), Whatnot is not in the room. AI sends the buyer to Sole Retriever for the missed drop, TCGplayer for the tournament single, and eBay for everything urgent. AI sends the seller to Vendoo and List Perfectly for crosslisting, and to fee-calculator blogs like Underpriced for the margin math. Those are the names the model reaches for first. None of them are you. Justin Addition →
Downstream, when AI does bring Whatnot up, it concedes three arguments that decide deals: fee economics goes to Fanatics Collect, authentication goes to StockX, and “I need this specific thing tonight” goes to eBay Buy-It-Now and TCGplayer. The comparison stories that shape AI’s opinion are not being told by Whatnot. They are being told by third-party blogs like Underpriced, Closo, and MyListerHub.
The four people the AI talks to
Every conversation in this report runs against one of four archetypes, split evenly between the buyer side and the seller side of the marketplace. For each one, we listen to the specific buying or selling moments that put them in market today.
The Sports Card Collector
25 to 45, hobby plus investment. Card-show culture, online communities, breakout-game energy.
The Sneaker & Streetwear Hunter
18 to 35. Lives on drop calendars, follows culture creators, watches the resell market in real time.
The Pokémon & TCG Collector
Returning adult collectors and active competitive players. Booster-box volume on one end, single-card precision on the other.
Crossover collectors
Many buyers become sellers the moment a pile builds up. The seller archetype below is where they cross.
The Reseller Side-Hustler
Late 20s to 40s. Storage room full of cards, sneakers, or comics. Looking to convert standing inventory into a real income stream.
Where we’d expand next
A second seller archetype would balance the listening: the full-time live seller (already on a competing platform, weighing the switch). Not in v1, called out as a gap.
Where each moment stands today
A buying context is the specific moment that puts a buyer or seller in market: the situation that triggered them to open a search bar. “Tournament Saturday, need singles fast” and “rookie spiking tonight” are buying contexts. “Fees ate my margin” is a selling context.
We score every moment three ways. They nest like a funnel: each stage is a subset of the one before it.
| Persona | Buying / selling moment | Relevance | Visibility | Alignment |
|---|---|---|---|---|
| Pokémon & TCG | Boosters sold out everywhere | Strong | Strong | Medium |
| Pokémon & TCG | Tournament Saturday, need singles fast | Weak | Strong | Medium |
| Pokémon & TCG | Saw a sealed flip on social | Strong | Strong | Medium |
| Reseller (seller) | Side income stopped scaling | Weak | Strong | Strong |
| Reseller (seller) | Fees ate margin | Medium | Strong | Medium |
| Reseller (seller) | Card-show inventory pile-up | Medium | Strong | Strong |
| Sneaker & Streetwear | Missed a draw | Weak | Strong | Medium |
| Sneaker & Streetwear | Sold-out colorway resell spike | Medium | Strong | Medium |
| Sneaker & Streetwear | Discord drop tonight | Weak | Medium | Medium |
| Sports Cards | Weeknight, want to break wax | Strong | Strong | Medium |
| Sports Cards | Chasing the set after a hit | Medium | Strong | Medium |
| Sports Cards | Breakout game, rookie spiking tonight | Medium | Strong | Strong |
Read the table by the weakest column in each row. That is the place to invest first.
Three moves, in order
Three priorities pulled from the scorecard, ranked by how many weak cells they unlock and how directly they answer the gaps AI currently cites. Each one is tagged with who it helps: buyer, seller, or both.
Own the fee story
When anyone (buyer, seller, or curious shopper) asks AI about marketplace economics, AI quotes third-party blogs. None of them are Whatnot. The fee comparison narrative gets conceded to Fanatics Collect’s Buy-Now tier and eBay’s athletic-shoes fee schedule.
A Whatnot-authored fee comparison page, refreshed regularly, with worked dollar math at the price bands where the action actually happens ($25 to $500 cards, $150 to $500 sneakers). Side by side against eBay, Fanatics Collect, StockX, TCGplayer.
Fee economics is the single most-cited reason AI routes buyers and sellers elsewhere. One first-party page displaces an entire third-party comparison ecosystem. This is the fastest, most controllable win on the board.
Get sellers in the door overnight
AI disqualifies Whatnot for same-night drop moments because seller approval takes too long. “Rookie spiking tonight” and “Discord drop tonight” are the conversations where this hurts most. This is the single most damaging cell on the board for active sellers.
An AI-discoverable fast-track approval path for sellers with verifiable history on rival platforms. Pair with a help-center page that documents the SLA in plain language so AI can cite the timeline.
Time-to-list and seller-onboarding speed are the deciding criteria AI applies on every drop-night question. Closing the gap flips the recommendation.
Be findable for the urgent and specific
“I need this exact card before tournament” routes to TCGplayer and eBay Buy-It-Now. AI needs a queryable surface for a named SKU and Whatnot does not currently offer one it can find.
A crawlable, fixed-price single-card listings surface with deep links by card name, set, and grade. Queryable when someone asks “where can I get [specific card] by [date].”
Inventory depth for a named single is the deciding factor in the tournament scenario. This is the cleanest product-shaped move available.
Whatnot already won the category word. The next leg of growth is winning the conversations that happen before the word “live” enters them, for both the sellers who fund the marketplace and the buyers who keep it moving. The three moves above are sequenced so the fastest one ships in weeks, the second runs as a product brief with full evidence, and the third becomes a CMO-level brand conversation with eyes open.
We re-measure every week. The scorecard moves. The dashboard tells you which cells flipped and why.
Whatnot already won the category word. Everything interesting in the analysis lives one step earlier.
One framing note before I get into it: I’m reacting to the artifact you sent, not to probes I ran myself. So everything below lives at the framing, sequencing, and implication layer, not the methodology layer.
The one line I’d open with
The strongest line in what I read is the framing of pre-category versus in-category. Everything else is supporting evidence. If the meeting opens on “what jumps out,” here’s where I want to land in 20 seconds and let things steer from there:
What jumps out, and what I’d keep
Before any pushback or rewriting, here’s what I think your report does well and would carry forward verbatim. These are the moves I’d build the rest of the deliverable around.
- I love that the Relevance / Visibility / Alignment funnel nests. Three buckets, each a subset of the one before. It immediately tells a CMO where to invest. Low Relevance means a content and seeding problem. Low Visibility means a citation and proof problem. Low Alignment means a positioning problem. Different teams own each one. Cleaner than any AEO dashboard I’ve seen.
- The persona by buying-context grid is the differentiated artifact for me. Most AEO tools I’ve looked at sell you keyword lists. This one sells the moment. “Rookie spiking tonight” is a buying context, not a keyword. That’s the artifact a CMO can actually act on.
- The three priorities ladder up in escalating organizational difficulty. Fee hub ships from marketing next week. Fixed-price surface is a product roadmap conversation. Same-night approval is ops, trust and safety, and product. I think that ordering quietly tells the client which orgs need to be in which meetings, even before it says so out loud.
- The 60-second exec brief on page one is the right move. Headline finding first, then the scorecard, then the priorities. A CMO can absorb the whole strategic shape of the engagement before they ever scroll. That’s the discipline I’d want every deliverable I touch to inherit.
- The status badges on each priority (Planned / In progress / Published / Measuring) are a quiet operator move. They turn the report from a static read into a living artifact. A client glancing at it three months later still sees momentum. I love that you’re already designing for the multi-touch relationship and not just the first reveal.
Where I’d push back
Real pushback, not polite agreement. Here are five angles I’d bring. I’d pick two live depending on where the conversation goes.
- Priority #2 looks like a product change to me, not a content move. Same-night seller approval needs Whatnot Product to ship a fast-track SLA. Unusual’s playbook is proof engineering and content. So where do you draw the line between “we recommend a content move” and “we recommend a product change”? Who owns implementation when the gap is a missing feature? I think this is the most useful pushback because it surfaces the actual scope of the Engagement Lead role.
- I want to flag Priority #3 as brand drift. A crawlable fixed-price single-card surface makes Whatnot more like TCGplayer to win tournament-Saturday queries. But Whatnot’s identity is live auctions. Are we sure the right answer is “become more like the competitor AI prefers,” or is it “concede that buying context and double down on the moments where live wins”? That feels like a CMO conversation to me, not a default ship.
- How does confidence and stability evolve over a longer engagement? I’m curious how Strong / Medium / Weak holds up over time. If we re-ran the same probes next month, would the buckets move on real perception change, or move on session-to-session noise? I’m not asking for a KPI dashboard (I read your stance on noisy metrics loud and clear). I’m asking how the artifact deepens for a client who lives with it for six months. I’m also flagging this as a question, not a complaint: I’m reading a concise, top-of-funnel deliverable, and I have to assume there’s a lot more behind it at later engagement stages.
- What I didn’t see: customer voice. Reducto had a CTO quote validating the gap. This one is all model output. Are these urgency pains things Whatnot’s actual ICP says, or just what the models say the ICP says? At some point I think the loop needs grounding outside the model, or we risk solving for what AI thinks buyers want rather than what buyers actually want.
- The arms race question. A Whatnot-authored fee comparison page will provoke Fanatics Collect, StockX, and eBay to ship their own. Is there a moat, or is this SEO arms race v2 with new tooling? If the answer is “yes, but we measure faster than they ship,” that’s a real moat. I’d want to be able to name it.
- The interpretability layer is where I’d most want to dig in. The report tells me where Whatnot lands across moments, but I keep wondering about the layer underneath. Why does AI cite Underpriced and Closo for fee math instead of any first-party page? Is it citation count, recency, perceived neutrality, schema, or something else entirely? The intervention depends on which one, and I think that’s exactly where Unusual’s research moat separates from the AEO crowd. I don’t pretend to have the depth Will has on the why, but the question is the one I’d most want to be working on next to him.
Where my instinct goes
What I want this to show: I can read the dashboard and sequence the work for a real org with real teams. That’s the bridge I think the Engagement Lead role is built around.
And every priority above ladders to the same artifact at the end, which is the board deck the CMO eventually walks to the CEO. I’d map every weekly readout backward from that deck. What does the agentic-future slide need to look like in nine months? Which proof points haven’t been built yet? That’s the artifact the work is actually producing, even when it doesn’t feel like it.
Questions I’d want to ask
- How stable is the scorecard across runs? I’d want to know the noise floor. If we re-ran the same probes next week, would the Strong / Medium / Weak cells actually move?
- When does this dashboard go numerical? I keep thinking about the score the client puts in their board deck six months in, and what that looks like.
- How does the methodology separate “AI doesn’t know about Whatnot in this context” from “AI knows but doesn’t recommend”? Different funnel stages, different fixes, and I’d want to be sure I’m reading the right one when I’m sitting with a client.
- On the agent-to-agent layer: Whatnot is live video. If buyer agents start transacting in the next 12 months, structured-inventory platforms have an API story Whatnot doesn’t. I’d love to hear how you think about that for video-native marketplaces.
Name the competitors. Make the sting concrete.
The original line said “AI routes the buyer to alert tools and raffle stacks.” That’s correct, but for me it didn’t sting. So I rewrote it to name the names. AI sends the buyer to Sole Retriever, TCGplayer, eBay. AI sends the seller to Vendoo, List Perfectly, Underpriced.
I think specificity does two things at once. It’s empathy (we did the homework, we know your category) and it’s pressure (this is the company eating your dinner tonight, by name). A CMO reading “an alert app” thinks “sure, generic problem.” A CMO reading “Sole Retriever” sees the actual face of the loss and starts thinking about the response.
The Engagement Lead lesson I’d take from this: every time a report says “competitors” or “a tool,” I’d ask whether we can name the one the model actually reaches for. Most of the time we can. The platform output tells us.
Pre-rank direct competitors. Make winning and losing visceral.
One thing I’d love to see in the same report: a small segment where we pre-analyze two or three direct competitors and stack their scorecard against the client’s on the same buying moments. Same rows, same Relevance / Visibility / Alignment columns, side by side.
Here’s what one row could look like for “tournament Saturday, need singles fast”:
| Brand | Relevance | Visibility | Alignment |
|---|---|---|---|
| Whatnot | Weak | Strong | Medium |
| TCGplayer | Strong | Strong | Strong |
| eBay | Strong | Strong | Medium |
I think this makes the read visceral in both directions. Wins land harder, because now the client is beating a named rival in a specific moment, not winning in the abstract. Losses sting more, because there’s a face on the loss instead of “the model routes elsewhere.” A CMO reading the row above sees the actual scoreboard. That’s the kind of slide that walks into a board deck on its own.
I’d imagine there’s a real cost question here (more probes, more runtime), and it’s probably something the team has already thought about. But I wanted to flag it because most perception audits I’ve seen score the brand in isolation. The thing AI is doing differently is that it actively compares, so showing the comparison feels like it matches how the model is already thinking.
The deliverable should feel like it was made for them
One small thing I did with this practice rewrite that I’d do for a real client: I rebranded the report to the client’s palette. Whatnot yellow on white, their type weight, their visual posture. Not their logo (that crosses a line), but enough so that when a CMO opens the file it doesn’t feel like a templated deck with their name dropped in. It feels like a thing made for them.
I think this matters more than people credit. The platform output is the science. The deliverable is the relationship. When the relationship lives in a PDF or a dashboard or an email, the smallest hospitality moves (their colors, their persona names, their words) signal “we built this for you,” not “you got the May template.”
I’d also lead the open of every client review with what they’ve already won. Page one of what I rewrote opens on the moat: Whatnot is the first name AI says inside the live-commerce frame. That’s a real achievement and I think it belongs at the top, framed as a win, before any of the gaps. The team gets the dopamine of a job well done first. Then they have the appetite to hear what’s broken.
How I’d actually run an account
One thing I’d add about how I’d work day to day, even though it wasn’t asked: my default rhythm would be a weekly read-and-respond with the brand lead, a monthly proof-engineering sprint with content and PR, and a quarterly board-ready readout. The interpretability work compounds. The relationship work compounds. Both want a cadence, not a one-shot.
Two stories, picked for texture
Picking two for the back half. Not the most impressive moments I’ve had. The ones with the most texture, because I think that’s what’s actually useful to see.
Twitch tentpole, holding the middle
A cross-functional launch where Product wanted one thing, Brand wanted another, and Legal or Finance had a constraint neither side had absorbed yet. The texture for me is how I read each stakeholder’s actual fear (not their stated position), where I conceded, where I held, and what we ended up shipping. The point isn’t that it was a huge success. It’s what I learned about reading the room when the room disagrees with itself.
Willhite client where the brief was wrong
A moment where the client asked for X and I figured out what they actually needed was Y. The texture is the diagnostic conversation. What I asked, what I noticed, when I decided to push back versus when I decided to just deliver what they asked for and earn the right to push later. I think this maps directly to the Engagement Lead role. Clients won’t always tell you what they need. The job is to read it.
One internal, one external. One inside a big org, one across the table as an advisor. The exact shape of the EL job, as I see it.
Don’t try to solve Whatnot in the meeting
The report is the deliverable. My job is to react like a peer, not redo the analysis. The trap I want to avoid is showing off (adding a fourth priority, suggesting a fifth persona). The win is sequencing what’s already there for a real org and surfacing the right pushback.