eBay's AI And Focus Verticals: Signal Vs. Reality For Operators And Investors
Semafor's CEO Signal profiled eBay CEO Jamie Iannone, crediting a focus-vertical strategy and AI for a stock turnaround. The pitch: go deep where enthusiasts spend, reduce friction with AI, and unlock the value of 134 million active buyers and decades of data.
The question for executives: do the operating metrics match the narrative? On several fronts, the story looks cleaner than the execution.
The Narrative: Focus Verticals + AI = Less Friction
eBay's thesis is simple: concentrate on enthusiast categories, apply AI to compress the time and effort between intent and transaction. The company highlights image-based listing, automated descriptions, pricing guidance, and buyer signals.
That positioning plays well on stage and with investors. But the core metric that should validate it-consistent buyer growth-has been moving the other direction.
The Numbers: Buyer Growth Hasn't Recovered
According to eBay's reporting, Q2 2025 marked the 13th consecutive quarter with fewer Active Buyers than in Q1 2018 (using the company's restated definitions). The small ~1% year-over-year bumps in recent quarters don't change the multi-year trend.
eBay now defines Active Buyers as anyone who completed a paid transaction in the last 12 months; multiple registrations per person are possible, and employee accounts are included if they transact. The company also restated GMV and Active Buyers at the end of 2021 back to 2018, which makes apples-to-apples trend analysis essential.
The "enthusiast buyers"-a strategic centerpiece-have been stuck at ~16 million since Q4 2022, down from ~17 million. Marketing has targeted these users, who eBay says drive ~70% of GMV, but aggressive discounts and feature tweaks haven't moved the number.
Does AI Actually Reduce Friction?
On paper, AI-driven listing flows should cut seller effort and increase data quality. In practice, many users report misidentified products, generic descriptions, and extra manual cleanup-especially in specialized categories where accuracy matters most.
There are also risks when AI summarizes or alters seller-provided content without clear disclosure. If buyers receive something that doesn't match the AI-edited description or images, you can expect avoidable disputes and support costs.
If you are operating a similar playbook, track these four metrics weekly: time-to-first-list from photo capture, list-to-sold conversion by category and condition, dispute rate tied to AI-edited content, and return rate due to "item not as described." If those don't improve, your AI isn't removing friction-it's moving it.
Org Design And Accountability: Where AI Lives Matters
Iannone highlighted personal use of AI for internal content creation and referenced a new "AI Design Technologist, CEO Office" role. That signals CEO-level sponsorship, which can help speed.
The risk: governance ambiguity if core AI responsibilities sit both in the CEO's office and under the Chief AI Officer. Without a clear RACI, you get duplicate tooling, conflicting standards, and uneven model quality. Centralize standards, decentralize experiments, and make content integrity a non-negotiable gate.
Cost-savings anecdotes (e.g., replacing agency work with AI-generated internal videos) raise a separate question: procurement discipline. Wins should be verified against prior actuals, not hypothetical budgets.
Customer Listening: Stories Vs. Systemic Signals
The profile cites CEO conversations leading to changes like more images for handbag listings and loyalty alerts. Useful, but incremental. Meanwhile, seller forums and unbiased user groups surface higher-impact pain-search quality, trust, returns, and listing accuracy.
Executives should require a standing "top five friction sources" list, tied to measurable outcomes and owned by cross-functional leaders. CEO listening tours are valuable; they shouldn't substitute for systematic signal.
What Leaders Can Learn From eBay's Moment
- Pick fewer metrics that matter: Active Buyers, conversion, repeat rate, and dispute/return rates. Make them visible and owner-assigned.
- Audit the AI funnel end to end: photo-in to funds-out. Find where humans rework AI output and fix at the model or UX level.
- Don't let "focus verticals" shrink your TAM: prove the cross-sell path and track cohort spillover into adjacent categories.
- Stop counting activity you can't scale: discounts that spike activity without improving lifetime value are noise.
- Protect content integrity: if AI edits seller assets, disclose it and give opt-outs. Measure the impact on claims and CSAT.
- Clarify ownership: keep AI platform standards under one leader; let business lines run controlled experiments on top.
- Replace anecdotes with baselines: every "we saved X" claim should link to a budget line and a before/after run-rate.
- Pressure-test narrative fit: when the story says "less friction," the data should show fewer clicks, faster list-to-live, and higher sell-through-by category.
What To Watch Next (Next 2-4 Quarters)
- Active Buyers and enthusiast-buyer counts trend vs. promotions spend.
- List-to-sold conversion and repeat purchase rates in focus verticals.
- Dispute and return rates tied to AI-generated descriptions or altered images.
- Org clarity: whether AI strategy centralizes under the CAIO or fragments across executive offices.
For primary sources and definitions, see eBay's investor relations disclosures and quarterly metrics.
eBay Investor Relations
Semafor
Building Your Own AI Operating Model?
If you're formalizing roles, controls, and measurement around AI, these curated paths can help you accelerate without creating chaos.
Bottom line: the pitch is crisp, but sustained buyer growth and cleaner conversion will be the proof. Until the metrics improve, treat the story as a hypothesis, not a conclusion.
Your membership also unlocks:
