When the purchase completes inside the AI, the conversion path is rewritten — and so is the case for earned authority. A formal examination of agentic checkout, the attribution and data consequences, and the link builder’s role in a zero-click commerce world.

Professional-advice note.  This article discusses commercial, contractual and data-protection considerations at a general, informational level. It is not legal, financial, tax or data-protection advice. Decisions about payment processing, customer-data handling, protocol contracts and regulatory compliance should be taken with qualified professional advisers familiar with your jurisdiction and circumstances.

For three decades, the conversion path in digital commerce has been broadly stable in shape: a prospective customer discovers a product, visits a website, browses, and completes a purchase on that site. Every discipline in marketing — analytics, attribution, retargeting, loyalty, conversion-rate optimisation — was built on the assumption that the transaction happens on a property the merchant controls and can observe. Agentic checkout dismantles that assumption. When a buyer completes a purchase inside ChatGPT, Gemini or another AI surface, the defining event of commerce moves off the merchant’s property entirely.

This is not a marginal development. By the most recent holiday season, AI agents were already driving a material share of orders — Salesforce data put it at roughly 20% of global orders, representing some $262 billion in sales. Klaviyo’s 2026 research found that 39% of consumers had purchased a product on the basis of an AI recommendation in the previous six months. Walmart, by one account, now receives around 36% of its referral traffic from ChatGPT. Whatever one’s view of the precise figures, the direction is unambiguous: a growing proportion of transactions is migrating into conversational surfaces, and the link building profession needs a considered position on what that means.

The purpose of this article is to provide that position. It is written for link builders and SEOs rather than payments engineers, so it treats the protocols and plumbing only to the depth required to reason about strategy. Its central argument is that agentic checkout, far from diminishing the relevance of earned authority, makes that authority more important — because in a world where the click disappears, the brand signals link builders cultivate become one of the few remaining levers a merchant genuinely controls.

The thesis of this article:  Agentic checkout collapses the conversion path and, with it, much of the data and relationship a merchant has historically owned. What survives the collapse is the brand — its recognition, its reputation, the web-wide authority that decides whether an agent selects it in the first place. The work of building that authority does not become less valuable when the click disappears; it becomes the principal way a brand defends itself against disintermediation.

What this article covers

• What agentic checkout is, and the current state of the protocols (ACP and UCP)
• How the conversion path is rewritten when the purchase completes inside the AI
• The data wall: attribution collapse, the loss of first-party data, and the ‘ghost supplier’ risk
• Why earned authority becomes more, not less, important in a zero-click commerce world
• Measurement when the click disappears: branded search, share of voice, and assisted signals
• A measured action framework for link builders and the brands they serve

What agentic checkout is

Agentic checkout refers to the completion of a purchase by, or within, an AI agent acting on a buyer’s behalf. Rather than directing the buyer to a storefront to transact, the agent either facilitates the transaction conversationally — collecting shipping and payment details and confirming the order within the chat — or hands the buyer to the merchant to complete the final step. The mechanism is governed by two open standards that emerged in early 2026.

What makes this more than a new checkout button is where it sits in the funnel. Agentic capability is moving steadily deeper, from early-funnel browsing assistance and product matching toward cart creation and secure payment. As it does, it consumes the stages of the journey one by one — discovery first, then consideration, then increasingly the transaction itself — and each stage it absorbs is a stage the merchant stops observing directly. The agentic layer effectively sits upstream of traditional SEO and paid acquisition: by the time conventional channels would have engaged the buyer, the agent has often already shaped the consideration set. Understanding agentic checkout therefore means understanding not just a payment mechanism but a reordering of the funnel, in which the merchant’s traditional points of influence are progressively enclosed within the conversation.

The Agentic Commerce Protocol (ACP), developed by OpenAI and Stripe and released under an open-source licence, defines how an AI agent, a buyer and a merchant collaborate to complete a purchase within a conversational environment. It is the native protocol behind ChatGPT’s Instant Checkout. The Universal Commerce Protocol (UCP), co-developed by Google and Shopify and launched at the start of 2026, is broader: it spans the full shopping lifecycle — discovery, cart, payment, fulfilment and post-purchase — across Google AI Mode, Gemini and other surfaces, and is endorsed by a substantial roster of major retailers.

A useful way to hold the distinction in mind is that UCP maximises the discovery surface while ACP optimises the conversational conversion. The most prepared retailers treat them as complementary rather than competing, and there is evidence the combination pays: merchants with dual implementation have been reported to capture meaningfully more agentic traffic than single-protocol stores. For the large population of Shopify merchants, the practical reality is simpler still — Shopify’s Agentic Storefronts abstract both protocols, so the merchant toggles on the AI channels they wish to sell through and the platform handles the underlying work.

It is worth being precise about the division of labour, because it explains why one protocol does not simply replace the other. UCP, being lifecycle-spanning and surface-agnostic, is concerned with making a merchant’s products discoverable and transactable across many interfaces — search, assistants, potentially smart-home and other emerging surfaces — and it preserves an important principle: the merchant remains the seller of record, so the underlying commercial and customer-data relationship is not handed wholesale to a marketplace. ACP, being specialised for the conversational transaction, concentrates on making the chat-to-purchase moment frictionless within environments like ChatGPT. A buyer might therefore be surfaced to via UCP when they ask a search-style question, and complete via ACP when the interaction is conversational. The two standards describe different segments of the same journey, which is why dual readiness, rather than a bet on one, is the prudent technical posture for merchants large enough to maintain it.

For everyone else, the platform layer increasingly hides this complexity. The merchant’s strategic attention is therefore better spent not on the protocols themselves — which are converging toward configuration rather than engineering — but on the question the protocols cannot answer: whether the agent chooses your product at all. No amount of protocol compliance compensates for not being selected, and selection is decided upstream of any checkout standard.

A moving target: the Instant Checkout trajectory

It is important to be candid about how new and unsettled this is. ChatGPT’s Instant Checkout moved into production for US users in early 2026, with Etsy as an early marketplace and a large pipeline of Shopify merchants onboarding. Within weeks, however, OpenAI signalled a recalibration, moving some purchase completion toward the merchant’s own apps and surfaces rather than keeping every transaction wholly in-chat. The lesson for strategists is not that agentic checkout is failing — adoption and order volume continued to grow — but that the precise locus of the transaction is still in flux. Building a strategy that depends on one specific checkout mechanism remaining unchanged would be imprudent.

Why the instability favours the link builder’s approach:  If the checkout mechanism itself is a moving target, investing heavily in any single transactional integration carries real risk of rework. The discovery and selection layer — whether an agent finds and trusts your brand enough to recommend it — is far more stable, because every version of agentic commerce, in-chat or redirected, begins with the agent choosing which products to surface. That selection is governed by authority, and authority is the durable investment.

How the conversion path is rewritten

Consider the classical conversion path and the agentic one side by side. In the classical path, discovery leads to a site visit, the visit generates observable behaviour, and the purchase completes where the merchant can see it, capture the customer’s details, and begin a relationship. In the agentic path, discovery, browsing, comparison and refinement all occur inside the AI, and the purchase may complete there too. The merchant receives an order. It does not receive the journey.

The structural consequence is that the messy, valuable middle of the funnel — where consideration happens and where brands have traditionally competed through experience and persuasion — increasingly takes place on a surface the merchant neither owns nor observes. The merchant’s influence is concentrated at two points: before the conversation, in whether the agent knows and trusts the brand enough to surface it; and after the transaction, in whether the brand can re-establish a relationship the platform has mediated. Both of those points are brand-authority problems before they are anything else.

A worked scenario: one purchase, two worlds

To make the abstraction concrete, follow a single buyer purchasing a £220 espresso machine, first in the classical world and then in the agentic one. The contrast is instructive precisely because the product, the price and the buyer are identical; only the path differs.

In the classical world, the buyer searches, clicks through to the merchant’s site, reads the product page, perhaps abandons and returns after a retargeting ad, then buys. The merchant observes every step: the referring keyword, the pages viewed, the time to purchase, the cart abandonment and recovery. It captures the buyer’s email at checkout, enrols them in a post-purchase sequence, and can later remarket accessories and beans. The transaction is the end of a visible journey and the start of an owned relationship.

In the agentic world, the buyer asks an assistant for a recommendation, discusses options conversationally, and completes the purchase through the agent. The merchant receives an order for one espresso machine. It does not see the comparison against three rivals, the questions about descaling and warranty, or the reason this product was chosen over the others. It may not capture a usable email, may not be able to remarket, and may find that the buyer’s sense of who solved their problem attaches to the assistant rather than to the brand. The same £220 changes hands; almost everything around it that the merchant once owned has evaporated.

What the scenario isolates:  The merchant’s leverage in the agentic world sits entirely at two moments the conversation does not contain: being selected as one of the candidate products in the first place, and being remembered afterwards. Both are won before and around the transaction, not within it — and both are functions of brand authority rather than on-site optimisation.

The strategic choice retailers now face

Beneath the tactical questions lies a genuine strategic decision, and retail leaders are making it openly: how much direct access to surrender in exchange for presence on the surfaces where buyers increasingly are. Three broad postures have emerged, and understanding them clarifies where link building fits in each.

The participation posture

Most multi-channel and DTC retailers are choosing to participate — enabling agentic discovery and checkout despite the data costs — on the reasoning that absence from the surfaces buyers use is more damaging than diminished visibility into the journey. For these brands, the priority is to participate while doing everything possible to remain the selected and remembered brand, which places earned authority at the centre of the strategy rather than the periphery.

The walled-garden posture

Amazon-first sellers operate largely within a single marketplace’s own agent and economics, optimising for that environment’s reviews, questions and pricing rather than for external agents. This is a different game with different rules, examined in this cluster’s dedicated marketplace material; the authority that matters there is concentrated inside the marketplace rather than across the open web.

The disintermediation-resistance posture

Some DTC brands see in agentic commerce an opportunity to reduce dependence on marketplaces by letting AI assistants route high-intent buyers more directly to them — but only if their data, authority and readiness are strong enough to be chosen. For these brands, investment in brand authority is not merely defensive; it is the mechanism by which they hope to bypass intermediaries altogether. The stronger the earned reputation, the more viable the strategy.

In all three postures, the common thread is that authority determines outcomes. The participant needs it to stay selected; the walled-garden seller needs its in-marketplace equivalent; the disintermediator needs it to win the direct relationship. There is no posture in which being widely trusted by the web is a disadvantage, and none in which a weak brand fares well.

The data wall: what the merchant loses

The most consequential effect of agentic checkout is not where the money changes hands but what the merchant ceases to see. Practitioners have begun to describe this as the ‘data wall’ or the ‘ghost supplier’ problem, and the concern is legitimate. When a purchase completes inside an AI surface, several assets that merchants have relied upon for decades may not reach them.

• First-party behavioural data. No site visit means no heatmaps, no browsing paths, no on-site experimentation — the raw material of conversion optimisation simply does not arrive.
• The customer relationship and contact details. Email capture, account creation and the permission to remarket may sit with the platform rather than the merchant, who receives the order but not the ongoing relationship.
• Attribution. With discovery and consideration inside the AI, conventional pixel-based attribution has a growing blind spot; revenue arrives without the journey context that explains it.
• Loyalty and lifetime value. If the buyer feels loyalty to the assistant that found the product rather than to the brand that made it, the merchant risks becoming an interchangeable supplier behind the agent.

This is the genuinely strategic risk in agentic commerce, and it is one that retail leaders are weighing openly: many are accepting a degree of lost direct access in exchange for presence on the surfaces where buyers increasingly are. The merchant remains the seller of record in UCP transactions, which preserves the underlying commercial relationship and the customer-data position more than a pure-marketplace model would; but the observable, usable richness of that relationship is diminished when the journey happens elsewhere.

The ‘ghost supplier’ framing deserves to be taken seriously rather than dismissed as rhetoric. If a buyer’s experience of solving their problem is mediated entirely by an assistant — the assistant understood the need, compared the options, and arranged the purchase — then the gratitude and the habit attach to the assistant, and the brand behind the product becomes a fungible component the agent could swap next time. A merchant in that position competes on whatever the agent optimises for, which may be price, availability or reliability rather than the brand equity it spent years building. The defence against fungibility is to be a brand the buyer specifically wants and the agent specifically trusts, such that substituting it would disappoint the user — and that, once again, is built through the recognition and reputation that earned authority creates. The data wall is real, but it is not equally damaging to all brands: it hurts unknown, interchangeable suppliers most, and recognised, trusted brands least.

Professional-advice note.  How customer data, consent and the seller-of-record relationship are treated varies by protocol, platform, contract and jurisdiction, and these terms are evolving. Treat the descriptions here as general orientation, and obtain specific guidance on data-protection obligations, consent flows and contractual terms before relying on any particular arrangement.

The reframing this forces:  If the journey and much of the relationship now happen off your property, the question is no longer only ‘how do I optimise my funnel?’ but ‘how do I remain the brand the agent chooses and the customer remembers?’ The first question is a CRO problem you can increasingly no longer fully address; the second is a brand-authority problem you can. The locus of competitive advantage shifts accordingly.

Why earned authority matters more, not less

It would be easy to read the data wall as bad news for everyone in search and conclude that the discipline is being automated away. That reading is too pessimistic and, more importantly, it misidentifies where value moves. When the controllable, on-site levers shrink, the levers that operate before and around the transaction become correspondingly more decisive — and those are precisely the levers earned authority pulls.

Three mechanisms make this concrete. First, selection. Every agentic transaction, regardless of where it completes, begins with the agent choosing which products to surface. That choice is governed by the brand’s recognition and the corroborating evidence the web provides about it — the recommendation factors examined elsewhere in this cluster. A brand the wider web vouches for is selected; a brand it does not is invisible, and an invisible brand cannot be bought through any checkout, however well integrated. Authority is the entry condition for participation.

Second, recall. In a zero-click world, discovery and purchase increasingly precede any direct contact, and the brand that is remembered is the brand that is sought out by name later. Branded search and direct navigation rise as recognition grows, even when no click is attributable to a specific AI mention. The work of becoming a recognised, frequently-cited entity — the work of link building and digital PR — is what builds that recall, and recall is a durable asset the platform cannot mediate away.

Third, defensibility. The authority signals that decide agentic selection — citations, reviews, community standing, brand entity strength — are earned and cannot be purchased outright, which is exactly what makes them defensible. A competitor cannot acquire a two-year citation footprint overnight, and a platform cannot easily strip a brand of recognition the whole web has conferred. In an environment where so much is being disintermediated, an earned reputation is one of the few things a brand can build that genuinely belongs to it.

None of this is new work. It is the established discipline pointed at a changed landscape. The tactic set in the 15 link building strategies that actually work in 2026 applies almost unchanged; what changes is the rationale. Where once the case for a citation in a category guide rested on rankings and referral traffic, it now also rests on agentic selection and brand recall — a broader and arguably stronger case than before.

A fourth mechanism: trust as a transaction-completion factor

There is a further, less obvious way authority pays in agentic commerce, and it sits at the transaction itself. Agents are reluctant to complete purchases that may fail or disappoint, because a failed transaction reflects badly on the assistant and erodes the user’s trust in it. An agent therefore favours merchants it has reason to believe are reliable — not only in the narrow sense of accurate inventory and pricing, but in the broader sense of being a reputable brand the wider web treats as trustworthy. Reputation, in other words, lowers the perceived risk of recommending and transacting with a merchant. A brand corroborated by reviews, editorial coverage and genuine community standing is a safer bet for the agent to act upon than an unknown quantity with an immaculate feed but no external footprint.

This closes an important loop. The same earned signals that win selection at the discovery stage also reduce friction at the completion stage, because both are expressions of the agent’s confidence that the merchant is real, reliable and well regarded. Authority is not merely a discovery tactic that becomes irrelevant once checkout begins; it is a trust signal that operates across the whole agentic journey, from the moment a product is considered to the moment a purchase is confirmed.

The cumulative case:  Authority decides whether you are selected, whether you are remembered, whether your position is defensible, and whether the agent trusts you enough to transact. Four distinct mechanisms, one underlying asset. It is difficult to identify any lever in agentic commerce with comparable breadth of effect — and it is the lever the link building discipline exists to pull.

Measurement when the click disappears

If the click is no longer the unit of measurement, the discipline must adopt new instruments. Conventional last-click attribution becomes progressively less informative as agentic and zero-click behaviour grows, and clinging to it produces a misleadingly bleak picture in which valuable brand-building work appears to do nothing. A more honest measurement framework reads a constellation of signals that move together.

1. Branded search and impressions. When a brand is mentioned more often in AI answers, more people subsequently search for it by name. Rising branded impressions in Search Console are often the first visible sign of growing recall — the data signature of recognition created upstream.
2. Direct traffic correlated with branded interest. If direct sessions rise in step with branded search interest, that points to genuine recall rather than tracking artefacts. The two moving together is the validation; direct rising while branded interest is flat usually indicates a measurement issue instead.
3. AI share of voice. Sampled visibility across the major engines — how often the brand is named, in what position, against which competitors — is the nearest equivalent to rank tracking in this world, and the most direct readout of whether authority work is moving selection.
4. Citation tracking. Logging which sources the engines cite turns measurement into an action list: the publications and pages you are not yet named in are your next targets.
5. AI-referred and assisted signals. Where AI does drive a click it often appears as direct or referral from chat domains and converts at a markedly higher rate; segment it and treat the quality multiplier as part of the value, accepting imperfect attribution throughout.

The throughline is a shift from measuring traffic to measuring visibility and recall. This is a difficult transition for organisations whose reporting and incentives are built on last-click revenue, and it is as much a governance challenge as an analytical one; boards will increasingly demand rigour about whether AI initiatives improve outcomes, and the teams that can articulate brand-led measurement will be better placed than those still counting only clicks. A sound technical foundation underpins all of it — an uncrawlable page cannot be retrieved, recommended or measured — which is why crawler access and rendering, covered in our guide to AI bot crawl optimisation, and the broader technical SEO foundation remain prerequisites rather than optional extras.

A practical way to operationalise this is to construct a small dashboard of the signals above and read them as a system rather than in isolation. No single metric is decisive in a non-deterministic, zero-click environment; the persuasive evidence is correlation across several. When AI share of voice rises, branded impressions climb, direct traffic grows in step, and AI-referred sessions convert above baseline, the combined movement is a credible account of brand-building working even though no single click can be traced to a single AI mention. Conversely, a fall in share of voice that precedes a softening in branded demand is an early-warning system conventional analytics would miss entirely. The discipline is to commit to this constellation of measures before the numbers are needed in a board meeting, so that the reporting frame is established and trusted rather than improvised defensively when last-click revenue looks flat. Establishing that frame is itself part of the link builder’s remit in 2026, because it is what allows authority work to be valued correctly rather than dismissed for failing a measurement model that no longer fits the world.

A measured action framework

The appropriate response to agentic checkout is neither alarm nor inertia but a deliberate rebalancing of effort toward the levers that survive the collapse of the click. The following framework is ordered by durability — the things least likely to be undone by the next protocol revision come first.

1. Invest in selection authority as the priority

Becoming a brand agents reliably select is the foundational work, and it is earned-media work: sustained citation acquisition in the guides and comparisons engines read, original research that becomes a cited source, genuine review depth, and consistent brand-entity signals. The reactive end of this — timely, citable assets — is covered in our newsjacking and reactive-PR playbook, and even classic placements such as a niche edit into an already-ranking guide can put a brand inside a source agents retrieve.

Treat this as a standing programme with a cadence, not a campaign with an end date. The reason is the compounding nature of the asset: each citation both contributes to selection today and deepens the brand’s recognition for tomorrow, so the returns accrue to consistency over time rather than to bursts of activity. A sensible operating rhythm is to maintain a continuous pipeline — identifying the guides and comparison sources that agents cite for your category, earning inclusion in them methodically, and refreshing the most important citable assets so they remain current enough for freshness-sensitive engines. The brands that will dominate agentic selection in two years are the ones running this programme now, because the footprint they accumulate cannot be replicated quickly by a competitor who starts later. In an environment where the transactional mechanics are converging toward parity, the earned-authority footprint is the differentiator that does not commoditise.

2. Build brand recall as deliberately as you build links

Because recall is what survives a zero-click journey, treat the memorability of the brand as a first-class objective: distinctive, frequently-cited assets, consistent naming, and the kind of recognisable presence that turns an AI mention into a later branded search. The principles that make a page a citable brand impression are the same ones examined in our analysis of link building for featured snippets and position zero, where the value is reframed from traffic capture to brand registration.

3. Maintain transactional readiness without over-committing

Ensure the brand is technically able to participate — accurate product data, consistent pricing and inventory, the relevant protocol enablement (often a configuration step on platforms like Shopify rather than bespoke engineering). Do this to a high standard, but resist pouring disproportionate resource into any single, possibly-transient checkout mechanism. Readiness is necessary; betting the strategy on one integration is not.

4. Plan to re-establish the relationship the platform mediates

Where the journey happens off-property, design deliberate ways to bring the customer into a direct relationship after the fact — through the product experience, packaging, post-purchase communication and genuine reasons to engage directly. This is a brand and operations question more than a link building one, but it is the natural complement to authority work and worth flagging, with the data and consent caveats noted above.

5. Reframe measurement and reporting

Move the organisation’s reporting toward visibility, share of voice, branded search and assisted signals, and educate stakeholders on why last-click revenue understates brand-building work in a zero-click world. The benchmark data that supports this reframing is collected in our living link building statistics for 2026.

The framework’s shape reflects the article’s argument: the durable, high-priority work is authority and recall, the necessary-but-bounded work is transactional readiness, and the connective tissue is a measurement model honest enough to value what the click no longer captures. Sector context matters too — the calculus differs for a considered B2B purchase versus a hyper-local one, and the dynamics we set out for a recruitment and HR-tech site or a wedding and hospitality supplier illustrate how the same principles localise, while cross-border considerations are addressed in our international link building guide.

Three misframings to avoid

Because agentic checkout is unsettling and new, it attracts a number of confident but mistaken conclusions. Three are worth addressing directly, because each leads to a poor allocation of effort.

Misframing one: ‘agentic checkout makes SEO and link building obsolete’

This conclusion mistakes the disappearance of the click for the disappearance of the work that earns selection. Agentic commerce does reduce the value of some click-dependent tactics, but it raises the value of the authority signals that decide which products an agent surfaces and trusts. The discipline is being reweighted, not retired; the brands that read it as an excuse to stop investing in earned authority are conceding the one advantage that still functions.

Misframing two: ‘this is a payments-integration problem’

Treating agentic checkout as primarily a matter of wiring up the right protocol misallocates attention to the part of the problem that platforms are rapidly commoditising. Integration matters and must be done competently, but it is increasingly a configuration step, and it is downstream of the decision that actually determines revenue: whether the agent chose your product. A flawless checkout integration on a brand no agent recommends transacts nothing.

Misframing three: ‘we can win it back later with retargeting’

The hope that the lost relationship can simply be reconstructed through conventional remarketing underestimates the data wall. If the journey was invisible and the contact details did not reach you, the inputs that retargeting depends upon may never have arrived. The realistic path to a direct relationship runs through being memorable enough to be sought out by name and through deliberate post-purchase re-engagement, not through pixels that were never set. This is why recall is a strategic objective rather than a nicety.

The common error beneath all three:  Each misframing assumes the controllable, mechanical parts of commerce are where the game is won, and treats brand authority as soft or secondary. Agentic commerce inverts that: the mechanical parts are being automated and commoditised, while the earned, human-judgment-resistant signals — reputation, recognition, trust — become the scarce and decisive resource.

Conclusion: the brand is what survives

Agentic checkout is a genuine structural change, not a passing novelty. It relocates the defining moment of commerce off the merchant’s property, takes much of the funnel into a surface the merchant cannot observe, and threatens to interpose a platform between brand and customer. These are serious consequences, and it would be a disservice to minimise them. Retailers are right to treat the transition with care, and to take qualified advice on the contractual, data and regulatory questions it raises.

But the conclusion that follows is not retreat; it is clarity about where value now sits. When the click disappears and the funnel goes dark, the merchant’s influence concentrates at the edges of the transaction — in being the brand the agent selects and the brand the customer remembers. Both are functions of earned authority, and earned authority is the link builder’s stock-in-trade. The discipline is not diminished by agentic commerce. Its rationale is sharpened by it: in a world that is disintermediating almost everything, a reputation the whole web has conferred is among the few assets a brand can build that cannot easily be taken away.

The practical instruction, then, is steady rather than dramatic. Keep the brand technically able to transact wherever buyers are, without over-investing in any single mechanism that may change. Pour the substantive effort into the authority and recall that decide whether the brand is chosen and remembered at all. And measure the work by what it actually produces in a zero-click world — visibility, share of voice, branded demand — rather than by a click that, increasingly, never comes. The brands that internalise this will find that agentic commerce, for all its disruption, rewards exactly the patient authority-building that good link building has always been.