What should marketers build now for the AI-search future?

The future of AI search is uncertain. That is not a reason to postpone investment.

The more useful question is which capabilities remain valuable across the widest range of plausible futures.

AEO Updates recently outlined four possible directions for the decision layer: a <strong>40% hybrid marketplace</strong>, a <strong>35% consideration operating system</strong>, a <strong>20% personal-agent economy</strong> and a <strong>5% slower or fragmented path</strong>. Those percentages are editorial scenario weights, not measured forecasts, market shares or budget allocations.[1]

The scenarios diverge most at the commercial edge. They differ on who hosts the conversation, where paid media appears, how much of the choice process AI controls and whether a transaction happens inside an AI surface.

They converge much earlier.

Across all four, marketers need accurate brand facts, sourceable claims, useful content, reliable product information and a measurement system that keeps unlike outcomes separate.

<div class='not-prose my-8 grid gap-4 sm:grid-cols-3'><div class='rounded-2xl border border-[#1674C4] bg-white p-5'><p class='font-mono text-xs uppercase tracking-[0.18em] text-[#1674C4]'>01 · Build now</p><p class='mt-3 font-serif text-xl text-[#15303C]'>Durable foundations</p><p class='mt-3 text-sm leading-6 text-[#516674]'>Evidence, useful content, product feeds and fixed-protocol measurement.</p></div><div class='rounded-2xl border border-[#148F86] bg-white p-5'><p class='font-mono text-xs uppercase tracking-[0.18em] text-[#148F86]'>02 · Prepare now</p><p class='mt-3 font-serif text-xl text-[#15303C]'>Reusable readiness</p><p class='mt-3 text-sm leading-6 text-[#516674]'>Conversational knowledge, product data ownership and API readiness.</p></div><div class='rounded-2xl border border-[#7652C8] bg-white p-5'><p class='font-mono text-xs uppercase tracking-[0.18em] text-[#7652C8]'>03 · Wait for a signal</p><p class='mt-3 font-serif text-xl text-[#15303C]'>Access-dependent bets</p><p class='mt-3 text-sm leading-6 text-[#516674]'>Platform agents, native checkout and agent-to-agent transactions.</p></div></div>

The goal is not to predict one winning platform. It is to avoid building the wrong layer too early.

<h2>Start with the durable overlap</h2>

The hybrid-marketplace, consideration-operating-system and personal-agent scenarios account for 95% of the current editorial weighting. Even the slower path still needs reliable websites, useful content and measurement. Across all four futures, the foundational overlap is close to universal.

In every scenario, AI systems still need to identify the brand, interpret its claims, retrieve supporting evidence and connect an offer to a user need. Marketers still need to know where the brand appears, whether it is recommended, what it is associated with and which sources support the answer.

The safest near-term investment is therefore not a custom shopping agent. It is the <strong>evidence and data layer</strong> that any future agent, search surface or recommendation system would need.

<h2>Tier 1: build now</h2>

Tier 1 contains capabilities that already improve today’s search, content, paid-media and measurement operations. They do not depend on a particular 2030 outcome.

<h3>Build an authoritative brand source of truth</h3>

A first-party site should state clearly what the company does, who its products are for, which problems they solve, what the brand can prove, how products differ and where current policy, price, availability and support information lives.

This is not a database for its own sake. It is a governed editorial and product layer that lets people, search engines and AI systems find the same current facts. Every meaningful claim should have an owner, source, status and review date.

<h3>Publish content that can survive retrieval</h3>

Crawlability and machine readability are necessary, but they do not create preference. The content also needs to answer real questions with specific, extractable evidence.

Useful formats include product pages with eligibility and limitation language, comparison pages built around decision criteria, original research with transparent denominators, attributable case studies, current policy pages and expert explanations that distinguish observation, interpretation and hypothesis.

AEO Updates’ prompt-family analysis showed why one polished answer is not enough. Measurement should test whether the evidence survives paraphrases, reruns, engines and modes.[2]

<h3>Make existing paid media AI-surface ready</h3>

Google already serves eligible ads around AI Overviews. Existing Search, Shopping and Performance Max campaigns can supply inventory, but the format and geography depend on the placement. Ads within AI Overviews are available in English in 12 named countries. Adjacent placements have broader eligibility.[3]

This does not create a separately targetable AI Overview campaign. Google does not currently allow advertisers to target only AI Overviews or obtain segmented reporting for in-overview serving.[3]

The no-regret work is familiar: keep Merchant Center feeds current, maintain accurate pricing and promotions, complete shipping and return information, use clear product descriptions and identifiers, and align creative and landing pages with the need expressed in the ad.

For advertisers already using Search or Shopping, this is good campaign and retail-data hygiene. It is not a promise of incremental AI-search return.

<h3>Use Preferred Sources only where it fits</h3>

Google Preferred Sources is a user-controlled personalisation feature. An eligible publisher or frequently updated editorial destination can invite readers to select its domain or subdomain. Google may then show fresh, relevant content from that source more often for that user.[4]

This can be useful for a publication with a following. It is not a universal brand requirement, ranking factor or AI trust signal.

<h3>Establish a V1 measurement baseline</h3>

The six-measure structure is adapted from the V1 operating model used for Dig One, a Dig Insights product. Dig Insights sponsors AEO Updates. The definitions below are editorial specifications for this framework, not an independent validation of Dig One or universal industry standards.

<div class='overflow-x-auto not-prose my-8'><table class='w-full min-w-[820px] border-collapse text-left font-sans text-sm'><thead><tr class='bg-[#15303C] text-[#F7FAF8]'><th class='px-4 py-3'>Measure</th><th class='px-4 py-3'>V1 definition</th><th class='px-4 py-3'>Boundary</th></tr></thead><tbody><tr class='border-b border-[#CAD7D3] bg-[#EAF1EF]'><td class='px-4 py-3 font-semibold'>AI Visibility</td><td class='px-4 py-3'>Eligible response runs with a brand appearance divided by all eligible runs.</td><td class='px-4 py-3'>Keep vendor composite scores separate.</td></tr><tr class='border-b border-[#CAD7D3]'><td class='px-4 py-3 font-semibold'>Share of Voice</td><td class='px-4 py-3'>Focal-brand mentions divided by qualifying mentions in a named competitor universe.</td><td class='px-4 py-3'>Not market share.</td></tr><tr class='border-b border-[#CAD7D3] bg-[#EAF1EF]'><td class='px-4 py-3 font-semibold'>Recommendation Rate</td><td class='px-4 py-3'>Eligible decision-stage runs with an affirmative recommendation.</td><td class='px-4 py-3'>Separate recommendation from neutral listing.</td></tr><tr class='border-b border-[#CAD7D3]'><td class='px-4 py-3 font-semibold'>Average Position</td><td class='px-4 py-3'>Mean textual mention position or mean citation position.</td><td class='px-4 py-3'>Choose one object and name it.</td></tr><tr class='border-b border-[#CAD7D3] bg-[#EAF1EF]'><td class='px-4 py-3 font-semibold'>Strongest Association</td><td class='px-4 py-3'>Most frequent predeclared coded association in the sample.</td><td class='px-4 py-3'>Descriptive signal, not proof of consumer belief.</td></tr><tr><td class='px-4 py-3 font-semibold'>Owned Citation Share</td><td class='px-4 py-3'>Citations to a versioned owned-domain registry divided by all citation events.</td><td class='px-4 py-3'>Also report incidence, unique pages and concentration.</td></tr></tbody></table></div>

These are operational definitions, not universal standards. Semrush defines AI Visibility as a 0–100 composite, while a raw presence rate answers a different question. Profound’s Average Position refers to textual mention order, while another system may report citation order.[13][14]

Every result should preserve the prompt register, intent stratum, geography, language, engine, model, mode, dates, reruns, competitor set and treatment of missing responses. Citation, mention and recommendation should not be pooled into one undifferentiated score.[2]

<h2>Tier 2: prepare now, but do not overbuild</h2>

Tier 2 contains capabilities that are increasingly useful but still depend on product access, market scope or a maturing commercial model.

<h3>Make the brand answer questions conversationally</h3>

A conversational interface exposes information gaps quickly. If product differences, pricing, eligibility, policies or evidence are difficult for an agent to explain, they are usually difficult for a person to evaluate too.

The platform-neutral work is to define approved answers, build a comparison and objection library, map claims to sources, identify escalation paths and record which answers vary by geography, customer type or product version.

Google Search Business Agent is already rolling out to qualifying U.S. ecommerce retailers. Customisation requires a U.S.-based store, a verified Merchant Center account, a claimed brand profile and at least 50 approved free listings.[5]

That makes activation a reasonable Tier 2 test for an eligible retailer. It does not make conversational agents universally available.

<h3>Prepare agent-ready product and policy data</h3>

Marketers should know which system owns product identifiers, current prices and promotions, inventory, shipping, returns, eligibility notices, customer support and compatibility rules.

Google’s Universal Commerce Protocol guidance requires participating merchants to maintain several of these fields and to validate real-time stock, order totals and shipping during checkout.[6]

The practical Tier 2 task is an <strong>API-readiness audit</strong>, not an immediate integration. Identify the system of record, data owner, update frequency and failure mode for every critical field.

<h3>Design pilot metrics before the platforms arrive</h3>

Proposed measures such as Agent Engagement Rate, Top Objections, Agent-Assisted Conversion, Agent Eligibility Rate and Data Completeness can guide a pilot. They are not standardized platform metrics. Each needs an event taxonomy, denominator, consent treatment and attribution rule before launch.

<h2>Tier 3: wait for a signal</h2>

Tier 3 contains capabilities that may become important but remain limited, allowlisted, experimental or operationally expensive.

<h3>AI Mode ad formats and direct offers</h3>

Google continues to describe several AI Mode commercial formats as tests, pilots or roadmap items.[7] A marketer should not build a dedicated operating model around them without access and a measurable use case.

The trigger should be explicit: the format is available in a priority market, the brand is eligible, the product category is supported, the team can state a learning objective and the reporting is sufficient to evaluate it.

<h3>ChatGPT Sponsored Agents</h3>

OpenAI’s Sponsored Agents are a selected-advertiser limited alpha in the United States. The user enters a separate, clearly labelled sponsored conversation after choosing an ad. OpenAI states that the sponsored conversation is separate from ChatGPT’s independent answer.[8][9]

That is a test signal, not a market-wide capability. Brands should prepare the underlying knowledge and governance now, but they should not fund a large custom build without access, measurement and an accountable commercial owner.

<h3>Business Agent for YouTube Ads</h3>

Google’s YouTube beta moves brand conversation into the media experience, but its current scope is narrow. The official beta requires a qualifying Demand Gen and Merchant Center campaign, a minimum daily budget, U.S. and English eligibility, and mobile in-stream placement. Shorts are excluded.[10]

AEO Updates treated the beta as a proof point for the hybrid and consideration scenarios, not a reason to change the scenario weights.[11]

<h3>Native checkout and agent-to-agent transactions</h3>

Google’s Universal Commerce Protocol is a real open protocol, and Universal Cart shows a plausible cross-surface transaction path. Current participation remains phased. The retailer remains seller or merchant of record in the documented flows.[6][12]

Full integration should wait until the brand is selected for a platform programme, the capability launches in a priority market, the team can demonstrate a near-term commercial use case and the commerce stack can meet the required reliability, security and governance standard.

<h2>How the tiers map to the four scenarios</h2>

<div class='overflow-x-auto not-prose my-8'><table class='w-full min-w-[760px] border-collapse text-left font-sans text-sm'><thead><tr class='bg-[#15303C] text-[#F7FAF8]'><th class='px-4 py-3'>Scenario</th><th class='px-4 py-3'>Editorial weight</th><th class='px-4 py-3'>Capabilities that matter most</th></tr></thead><tbody><tr class='border-b border-[#CAD7D3] bg-[#EAF1EF]'><td class='px-4 py-3 font-semibold'>Hybrid marketplace</td><td class='px-4 py-3'>40%</td><td class='px-4 py-3'>Evidence, feeds, paid-media hygiene, conversational readiness and platform-specific pilots.</td></tr><tr class='border-b border-[#CAD7D3]'><td class='px-4 py-3 font-semibold'>Consideration operating system</td><td class='px-4 py-3'>35%</td><td class='px-4 py-3'>Brand source of truth, comparison content, recommendation measurement and association clarity.</td></tr><tr class='border-b border-[#CAD7D3] bg-[#EAF1EF]'><td class='px-4 py-3 font-semibold'>Personal agents become the buyer</td><td class='px-4 py-3'>20%</td><td class='px-4 py-3'>Structured product and policy data, APIs, eligibility, governance and transaction reliability.</td></tr><tr><td class='px-4 py-3 font-semibold'>Slower or fragmented adoption</td><td class='px-4 py-3'>5%</td><td class='px-4 py-3'>Strong websites, useful content, search operations and measurement still retain value.</td></tr></tbody></table></div>

These weights should guide strategic attention, not be copied into a budget formula. The framework deliberately overweights durable foundations because they remain useful under every scenario.

<h2>The investment sequence</h2>

<div class='not-prose my-8 rounded-2xl border border-[#15303C] bg-[#15303C] p-6 text-[#F7FAF8]'><p class='font-mono text-xs uppercase tracking-[0.18em] text-[#FF744A]'>Build now</p><p class='mt-3 leading-7'>Governed brand and product facts; source-backed claims; current paid-search and product-feed hygiene; fixed-protocol measurement; optional Preferred Sources promotion for eligible editorial destinations.</p><p class='mt-6 font-mono text-xs uppercase tracking-[0.18em] text-[#FF744A]'>Prepare now</p><p class='mt-3 leading-7'>Conversational answers and escalation; product, policy and support ownership; an API-readiness audit; eligible Search Business Agent testing; pilot event and attribution rules.</p><p class='mt-6 font-mono text-xs uppercase tracking-[0.18em] text-[#FF744A]'>Wait for a signal</p><p class='mt-3 leading-7'>AI Mode commercial formats still in test; ChatGPT Sponsored Agents; YouTube Business Agent; native or embedded UCP checkout; custom merchant agents and agent-to-agent transactions.</p></div>

<h2>AEO Updates Takeaway</h2>

The strategic mistake is not choosing the wrong 2030 scenario.

It is spending heavily on the commercial edge before the evidence and data foundation is ready.

Build the capabilities that every plausible future needs. Prepare the next layer while the access rules and interfaces mature. Fund the expensive agent and transaction layer only when a real platform, market or measurement signal arrives.

That is not waiting. It is sequencing the investment so that uncertainty becomes manageable rather than paralysing.

<h3>References</h3>

[1] <a href='/articles/who-owns-ai-decision-layer-2030-scenarios'>AEO Updates, Who will own the decision layer?</a>, 23 September 2026; scenario weights corrected 24 September 2026.

[2] <a href='/articles/prompt-not-market-ai-visibility-measurement'>AEO Updates, The prompt is not the market</a>, 30 August 2026.

[3] <a href='https://support.google.com/google-ads/answer/16297775?hl=en' target='_blank' rel='noopener noreferrer'>Google Ads Help, About ads and AI Overviews</a>, accessed 25 September 2026.

[4] <a href='https://developers.google.com/search/docs/appearance/preferred-sources' target='_blank' rel='noopener noreferrer'>Google Search Central, Help your readers find your site through Preferred Sources</a>, updated 18 September 2026.

[5] <a href='https://support.google.com/merchants/answer/16410382?hl=en' target='_blank' rel='noopener noreferrer'>Google Merchant Center Help, Get started with Business Agent</a>, accessed 25 September 2026.

[6] <a href='https://developers.google.com/merchant/ucp/guides/merchant-center' target='_blank' rel='noopener noreferrer'>Google Universal Commerce Protocol Guide, Prepare your Merchant Center account</a>, updated 25 August 2026.

[7] <a href='https://blog.google/products/ads-commerce/google-marketing-live-search-ads/' target='_blank' rel='noopener noreferrer'>Google Ads, A new generation of ads for the AI era of Search</a>, 20 May 2026.

[8] <a href='https://openai.com/index/reimagining-advertising-with-ai/' target='_blank' rel='noopener noreferrer'>OpenAI, Reimagining advertising with AI</a>, 16 September 2026.

[9] <a href='https://help.openai.com/en/articles/20001524-sponsored-agents-in-chatgpt-ads' target='_blank' rel='noopener noreferrer'>OpenAI Help, Sponsored Agents in ChatGPT Ads</a>, accessed 25 September 2026.

[10] <a href='https://docs.google.com/forms/d/e/1FAIpQLSdiWVBHjjMKeNgU9i3nfjP4LUTNz-RA7bM9T2TDcjkl6ZJemw/viewform' target='_blank' rel='noopener noreferrer'>Google, Business Agent for YouTube Ads beta interest form</a>, accessed 25 September 2026.

[11] <a href='/articles/google-business-agent-youtube-ads-decision-layer-proof-point'>AEO Updates, Google brings Business Agent to YouTube ads</a>, 24 September 2026.

[12] <a href='https://blog.google/products-and-platforms/products/shopping/google-shopping-cart' target='_blank' rel='noopener noreferrer'>Google Shopping, Introducing the Universal Cart</a>, 19 May 2026.

[13] <a href='https://www.semrush.com/kb/1594-ai-seo-metrics' target='_blank' rel='noopener noreferrer'>Semrush, AI Visibility Metrics</a>, accessed 25 September 2026.

[14] <a href='https://www.tryprofound.com/blog/how-to-track-your-visibility-in-ai-search' target='_blank' rel='noopener noreferrer'>Profound, How to track your brand visibility in AI search</a>, 26 January 2026.