OpenAI just released “Buy it in ChatGPT”, the first big shift towards conversational commerce. US can already try it out and make purchases. Etsy already there, Shopify coming soon.
Since the announcement, I’ve been thinking about what this means and what this changes, especially when you take into consideration also the Agentic Commerce Protocol they also launched and announced. So, we can buy things in chat, agents can talk to businesses that have shops about buying things, and very likely agents can also talk with other agents about buying things. People can also build agents that will under certain conditions buy some things.
And i’ve been writing down thoughts of what can and will change, what implications does this have on our experience but also on incumbents and what new things will this bring to the online shopping table. After all, this seems to be a completely new playing field or marketplace of sorts and at the very least we will see an inevitable wild west phase while this space develops and changes.
And how could it not change? For decades online shopping has been about search results, SEO, ads and clicks. Now it might shift into conversational commerce or any other choice of term you prefer.
By the way, I mentioned that I started writing things down. I got to over 100 ideas and perspectives of what might change, evolve or at the very least be influenced or affected by this new growing marketplace. The more I wrote, the more layers of disruption seemed to appear. Trust, curation, pricing strategies, how reviews will work, how ads will work, how will loyalty manifest, gifting, all of them will be influenced to some degree. Some of it will be revolutionary, some of it will go wrong and some will be completely unexpected. And my huge popcorn bag and I will be right here trying things out and enjoying the show.
In the meantime however, here’s the list of 100 perspectives and ideas to consider. My aim isn’t to predict the future but to keep a very broad and open mindset to what might be coming. Consider it a map of possibilities. And please, definitely leave a comment if I’ve missed any unique or important shifts.
1. Recency of Data & Price
- Real-time pricing feeds become essential – shops must provide always-updated data or risk being filtered out by LLMs that prioritize freshness.
- Dynamic pricing intensifies – LLMs may compare in real time, pushing shops into algorithmic pricing wars.
- New Data Kings – ChatGPT has already mentioned that we can use the card on file. In the long term, LLMs will have access to both our purchasing profile and history and reasoning behind the purchases as well as the transactions and cards on file.
2. How Would Trust & Reviews Work With “Buy In ChatGPT”
- LLM-native review aggregation – no one will scroll through 500 reviews; the LLM will summarize sentiments, or will be asked to do so
- Trust scoring systems – new rating systems may emerge that merge sentiment analysis, authenticity checks, and verified buyers. There might even be a rating for the “buy it in ChatGPT” experience.
- Fake reviews arms race – manipulation will shift from writing fake reviews (and even content) to manipulate the LLM summarization layer.
- “Explain this rating” prompts – buyers will interrogate the LLM to justify why an item is “trusted” or “popular”. Or investigate the shop itself for any issues as a default step of the buying experience rather than the exception.
- Conversation-based customer service reviews – LLMs could use transcripts of past interactions to judge reliability of sellers.
- Review for reward automations – The LLM can prompt: “Leave a 30-sec voice review now for 5% off next time?” and enforce authenticity via proof-of-purchase, time-to-review windows, and anti-gaming checks.
3. “Buy it in ChatGPT” Curation & Discovery
- LLM curation replaces search results – instead of “page 1 of Google,” discovery happens through conversation flow.
- Ads disguised as recommendations – new ethical battles: when the LLM “recommends,” is it paid influence or organic?
- Curation algorithms trained on taste profiles – personal taste embeddings may guide product surfacing.
- Hyper-personalized gift guides – “find me a gift for my 12-year-old niece who likes robotics” → LLM curates.
- Sponsored conversation turns – brands might “sponsor” the examples LLM uses in a dialogue.
4. New Roles & Jobs
- Prompt-SEO managers – specialists in optimizing how products are described to appear in LLM responses. Conversations or specific prompt phrasings related to the product or the job it does for a customer might get embedded in the product description or metadata.
- LLM trust auditors – ensuring shops aren’t unfairly suppressed or manipulated.
- Conversation designers for commerce – shaping how shops “speak” through LLM personas.
- Product embedding engineers – optimizing metadata so items are represented accurately in LLM vector spaces.
- AI marketplace regulators – new watchdog roles around fairness, transparency, and monopoly risk.
5. Impact on Existing Monopolies
- Amazon might lose its “search moat” – if discovery is conversational, Amazon’s endless catalog is no longer the main entry point.
- Etsy’s differentiation weakens – unless it ensures ACP feeds capture uniqueness, it risks being flattened into “just another store.”
- Platform lock-in shifts – instead of “shop on Amazon,” people will shop “through ChatGPT.”
- Opportunities for niche shops – smaller shops can bypass SEO and Google Ads by integrating directly into conversational flows in unique and novel ways.
6. SEO, Content & Pricing Strategies
- SEO loses relevance, Prompt-Optimization rises – content for Google becomes less relevant; descriptions optimized for LLM parsing matter more.
- Prompt equivalents of keywords – A new discipline emerges: Prompt SERP – optimizing product copy to match how people ask (“eco, hand-made, dad-joke vibe”).
- Conversational keyword stuffing – product listings may include phrases anticipating how users ask questions.
- Comparison requests replace search queries – “Compare two bow ties” becomes the norm. Shops must prepare structured comparison data.
- Long-tail content shifts to conversational knowledge bases – e.g., “What’s the difference between batwing and butterfly bow ties?” answered with embedded shop links.
- Long-tail “intent lexicons” – Brands build libraries of user intents (e.g., “quiet luxury,” “dopamine dressing”) mapped to products with evidence snippets the LLM can quote.
- Policy aware safeguards in reranking – Guardrails (no illicit items) merge with relevance scoring. The LLM must explain: “I’m excluding X due to policy Y—here are compliant alternatives.”
- Legal right to explainability – Expect regulations requiring “meaningful explanation” for why X ranked over Y – including which ads or incentives influenced the answer.
7. Buy It In ChatGPT – Discounts & Shop Communication
- Conversational promotions – “Do you have a discount code for me?” will be a standard user prompt.
- Buyer prompt-hacking for discounts – Consumers will learn scripts (“bundle + loyalty + price-match + cart-abandon analogues”). Shops might respond with transparent negotiation policies and rate limits. They might develop AI based negotiating agents too.
- Dynamic negotiation via LLM – haggling could be automated (LLM: “Can you do 10% off if I buy two?”).
- Personalized bundling – shops may surface bundle offers based on conversational context (“matching dad and son bow ties”).
- Loyalty rewards embedded – LLMs might track buyer history and suggest loyalty perks automatically.
- Contextual timed offers – “Since you’re buying for a wedding, here’s a 24h deal on cufflinks.”
- LLM-mediated seller engagement – Sellers can message through the LLM: follow-up sizing help, fabric care tips, matching items, all this without ever getting your email or phone, with strict user-controlled privacy gates.
- Cross-thread service mapping – The LLM links your support chats (“broken clasp last May”) to the original purchase, warranty, seller policy, and shipping events—making returns/repairs one prompt away.
8. Shops Adapting to LLM-Native Commerce
- LLM-ready product feeds – ACP-compatible data structures become as important as having a mobile-friendly site.
- Conversational brand voice – shops must define how their “AI agent” speaks in ChatGPT and must have one as table stakes in the future.
- Explainer assets – rich content (videos, diagrams, stories) that LLMs can pull into conversations.
- Narrative-driven shopping – items could be surfaced in story-like formats instead of lists.
- Full-stack ACP storefronts – shops may emerge with no website, only LLM integration.
- Culture engineering (meta content) – Brands and creators produce memes, micro-songs, and short scripts designed to become quotable snippets the LLM cites as “social proof” in recommendations.
- Micromarket pop-ups – Temporary “festival” or “wedding season” micro-malls spawn inside the chat, aggregating context-specific bundles across many sellers for a limited time.
- Location-Aware Shopping Prompter – LLMs can proactively surface reminders or offers when your physical location intersects with your shopping intent.
9. Potentially Revolutionary / Wrong / Unexpected
- Conversational cart collapse – no one “browses,” they just buy the first few surfaced items → monopoly by ranking.
- Bias lawsuits – if LLMs suppress certain sellers, antitrust and discrimination cases may explode.
- Black market prompt injection – shady actors could pay to hijack LLM conversations with hidden instructions so LLMs rank products higher.
- Overtrust by users – people may blindly accept LLM-curated items, leading to homogenized consumption.
- Unexpected resurgence of local shops – LLMs could surface hyperlocal inventory (“get this in 30 minutes from a nearby shop”).
- Regulatory crackdowns – governments may mandate transparency: “This recommendation was sponsored.”
- Conversational fatigue – some people may miss the “visual browsing” and not trust AI-curated shopping.
- Payment fraud scaling – attackers could exploit ACP to spoof shops or intercept transactions.
- Over-customization backlash – users may feel manipulated when LLMs “know too much” about their intent.
- New gray markets – LLMs surfacing “cheap but risky” sellers could undercut regulated marketplaces.
- Prompt-in-image adversarial hacks – Shops may attempt stealthy prompt injections inside product images. Expect scanning/stripping by platforms, watermarking, and penalties for offenders.
- LLM hygiene scoring for shops – Shops get ranked on feed freshness, policy compliance, safety, return reliability, and responsiveness, in turn affecting their “conversation prominence.”
- Personal Shopping Sandboxing – You can create experimental “alt profiles” to explore new styles without polluting your main taste model and then merge if you like the direction.
10. Experience & Interfaces
- Conversational receipts – instead of email, you’ll chat “Where’s my order?” and get an AI-generated update. Commercial conversations with transactions might also be saved differently to have proof later on.
- Shoppable chat threads – your conversation history doubles as your shopping history.
- Multi-agent marketplaces – multiple shop agents may “pitch” in the same conversation, like vendors in a bazaar. You get quotes from agents.
- Agent-to-agent bargaining – Your LLM negotiates with shop agents on bundle, shipping, and warranty; standardized ACP negotiation schemas emerge (with audit logs for fairness and review).
- Visual + conversational hybrids – LLMs will show curated galleries inline with conversational context.
- AI concierges for events – “Plan a wedding outfit” → LLM curates, compares, bundles across multiple shops.
- Zero-hop support from the purchase thread – You re-open the original convo: “My clasp snapped,” and the LLM spins up the seller’s AI or human support with order IDs, photos, warranty, and suggested remedies already attached.
- Purchase memory to anticipate buying – LLMs that remember what you bought (sizes, materials, colors, brands, price tolerance) start proactively flagging replenishments, upgrades, and matching accessories e.g., “Your kid’s size likely changed; want to auto-queue the next bow tie before the school event?”
- Contextual intent shaping from history – Past conversations (wedding, cosplay, minimalism, sustainability) become features in your “taste embedding,” so future recommendations match not just brand/price – but vibe and values.
- Holiday-aware wishlisting – You’ll say, “Keep an eye on gifts for Dad’s birthday and Black Friday,” and the LLM sets watchlists, reminder nudges, and early-bird bundles across shops—automatically checking stock and price drops.
- “Think of what I want” mode – You can instruct: “Curate surprises aligned with my style and constraints,” letting the LLM keep a running shortlist that evolves as your tastes drift or budgets change.
- Fine-grained “shopping moments” – The system discerns if you’re browsing, comparing, price-hunting, gifting, or urgent-buying and then adapts curation, detail level, and discount strategies depending on your state.
- Work vs. personal account separation – You’ll route purchases through profiles with separate budgets, approvals, tax receipts, and data silos – so personal wedding attire doesn’t pollute enterprise procurement.
- Ecosystem vs. single-hub UX – Either ChatGPT becomes the main mall or new GenAI-native malls emerge (visual canvases + multi-agent pitches), with ChatGPT/others providing the underlying reasoning APIs.
- Try on via image LLMs – “Dress me like this Pinterest board” becomes actionable: the model composes outfits from multi-shop inventory, simulates fit, and re-ranks picks based on how they actually look on you.
- Gift graph building – The LLM tracks who you gift, their sizes/interests, and avoids repeats, surfacing timely, non-duplicative ideas with auto-thanks and gift receipts.
- Human-in-the-loop escalations – For expensive or risky buys, the LLM defaults to “bring in a human concierge,” but keeps the conversation context and structured pros/cons for fast decisions.
11. Payments
- Native wallet integration – OpenAI or others may launch wallets to simplify ACP purchases. They already mentioned you can use the card on file in your ChatGPT account when you “buy it in ChatGPT”.
- Crypto-native transactions – direct AI-to-AI commerce using stablecoins could appear.
- Split payments conversationally – “Pay half now, half in 30 days” may be negotiated with the LLM.
- Biometric conversational auth – “Confirm purchase with your voice or Face ID.”
- User-set spending allowances – “Auto-approve under €30/day; require confirmation above.” The agent adheres to thresholds, with audit trails, anomaly alerts, and family/teams sharing controls.
12. Ads & The Ad Business
- Prompt-ads marketplace – brands bid to appear in conversational results.
- Sponsored conversational “turns” – subtle nudges (“Some buyers also loved this eco-friendly alternative”).
- Pay per conversation models – instead of pay-per-click, brands pay per meaningful interaction.
- Conversational retargeting – LLMs may reintroduce abandoned carts days later during unrelated chats, especially if they have their calendars or goals also integrated with the LLM.
- New walled gardens – OpenAI, Anthropic, or Google could dominate adtech in conversational commerce.
- Ethical ad disclosure battles – debates over how clearly “this is sponsored” should be marked in conversation.
- Ad-funded premium – Some LLMs may offer advanced shopping features free or a discount on your monthly subscription if you accept sponsor insertions in ranked suggestions – requiring very clear disclosures and opt-outs.
- Ad measurement reimagined – Attribution becomes “conversation-path based”: which turn, claim, or evidence snippet moved you from exploration to purchase. New analytics and ad pricing follow.
13. Other Considerations
- Environmental impact – LLM-driven commerce may increase energy costs for trivial shopping queries.
- Legal disputes over liability – if the LLM recommends a scam seller, who’s responsible?
- Rise of conversational affiliate marketing – influencers train “their LLM” to recommend affiliate-linked products.
- Shift in loyalty – buyers may be loyal to the LLM brand (ChatGPT, Claude) rather than the shop.
- Cultural homogenization vs. niche revival – either everything flattens to “most popular,” or micro-niches thrive through personalization.
- Ethics & bias councils for recs – Independent auditors certify recommendation fairness, source diversity, and ad labeling; badges influence ranking. Expect audits on sensitivity (e.g., weight-loss products, kid-targeted upsells).
- User-controlled portability – Data export of “taste embeddings + purchase memory + allow/deny lists” to switch LLMs, possibly even sell or lease your profile to lower subscription costs.
- Consentful data layering – You’ll toggle which signals the LLM may use (location, calendar holidays, email receipts). Tighter consent loops might boost trust and may increase willingness to share.
- Personal carbon budgets – You’ll set sustainability constraints; the LLM tracks embodied carbon per basket and suggests lower impact swaps or consolidated shipping.
- Crisis modes – During supply shocks or recalls, LLMs shift to reliability/safety-first ranking, pausing aggressive upselling and pushing verified stock with transparent ETAs.
Looking to spark change in your product journey? Here’s how I can support you:
- Product Management & Leadership Coaching – I help leaders grow their confidence, lead despite uncertainty, focus on delivering value and create real impact for their teams, users and company.
- Outcome Thinking & OKRs – Backed by hands-on experience in both using and rolling out OKRs across multiple teams and companies, I help you roll out, improve or fix your OKRs or OKR system and make changes and improvements that stick beyond the start of the cycle.
- Facilitation With a Twist – I create, facilitate and moderate workshops that energize the room and foster stronger engagement and learning – all through my crazy, fun and playful approach.
- GenAI for Product Teams – Practical, hands-on workshops that give teams and leaders the skills to harness AI and GenAI for their everyday goals and way of working, but also the time to reflect as a group on the impact that this AI wave brings.