Instinct’s new product recommendations are giving some users the ick

AI startup Instinct launched Instinct Selections, a human-curated recommendation feature spanning dining, travel and shopping, potentially creating advertising, affiliate or commerce revenue opportunities. Early user feedback was negative because unsolicited product suggestions appeared disconnected from users' needs, raising risks of spam perceptions and engagement pressure as Instinct competes with larger AI platforms. The rollout follows a $1 billion Series C at a $10 billion valuation; management says the agent is approaching $1 billion in annualized transaction volume, with travel accounting for more than 50%.

Analysis

The core issue is not recommendation quality but permissioning: unsolicited commercial inference from email, travel plans, or other sensitive context can raise churn before it produces meaningful take rate. For a consumer agent without disclosed active-user scale, this is a negative signal on monetization maturity; affiliate revenue is economically immaterial unless recommendations occur in high-intent sessions and conversion attribution is credible. The 1-3 month catalyst is whether the product is quickly moved behind an opt-in setting and whether management discloses conversion, repeat usage, or merchant-funded economics rather than transaction-volume claims.

META is modestly advantaged at the margin because its ad stack, commerce graph, and ranking infrastructure make sponsored recommendations a more natural extension of existing user expectations. But the second-order risk applies to META as well: agents that surface purchases from private conversational context face a higher privacy threshold than feed advertising, potentially limiting ad-load expansion and increasing regulatory scrutiny in Europe. Near term this is not earnings-relevant for META; it is a product-design datapoint supporting the view that AI-agent monetization will be slower and more trust-constrained than current narrative multiples imply.

The contrarian view is that initial backlash may be an acceptable cost if the agent can demonstrate superior travel conversion, where intent is explicit and basket values are high. A successful opt-in travel workflow could pressure OTAs and metasearch over 6-18 months, particularly EXPE and BKNG, but there is insufficient evidence on user scale, supplier economics, or whether bookings are incremental to support a position today.

Key Decisions for Investors

• No standalone trade on this development; treat it as a watch item rather than a valuation catalyst because active users, recommendation frequency, conversion, and monetization terms are undisclosed.
• Maintain META as the listed relative beneficiary of agent-commerce monetization, but do not add solely on this signal. Reassess after META reports AI-driven engagement, click-to-message, or business-messaging monetization; a disclosure that privacy constraints limit use of conversational context would falsify the upside read-through.
• Set a 1-3 month alert for an opt-in redesign, named commerce partners, and disclosed travel-booking conversion. Evidence of high-intent, opt-in conversion would justify reviewing a long META versus short EXPE basket; absent those data, OTA disintermediation remains speculative.
• For AI-exposed growth baskets, avoid underwriting near-term revenue from agent recommendations. Require evidence of merchant-funded take rates and stable retention before assigning more than de minimis value to commerce optionality; persistent unsolicited prompts or a public privacy backlash would be a negative retention signal.