AI Search Visibility: The Complete 2026 Guide

A skincare brand we worked with last quarter had a problem they could not see. Their organic traffic was steady, their Google rankings were fine, and their ad spend was efficient. Then their revenue from new customers started drifting down, quietly, two to three percent a month. It took a full audit before anyone realized the leak: when shoppers asked ChatGPT and Perplexity for a “gentle vitamin C serum under fifty dollars,” a competitor came up in the recommendation and they did not. Their store was invisible to the machines doing the actual recommending. That is the exact failure mode this guide exists to fix, because AI search visibility is now the difference between being in the consideration set and never being mentioned at all.

We ship answer engine optimization work every week for merchants and agencies, and the pattern is consistent. The brands winning in AI search results are not the ones with the biggest content budgets. They are the ones whose data is structured, retrievable, and machine-legible, so that when an LLM assembles an answer, their product or page is the easiest correct thing to cite. This is a systems problem more than a copywriting problem, and it rewards teams who treat their catalog and content as an API for agents rather than a webpage for humans.

TL;DR

What AI search visibility actually means: It is your likelihood of being retrieved, cited, and recommended inside AI-generated answers from tools like ChatGPT, Perplexity, Gemini, and Google AI Overviews, which is a different game from ranking blue links on a results page.
The three levers that move the needle: Machine-readable structured data, entity clarity and consistency across the web, and real-time accurate product or service data that agents can trust enough to act on, all reinforced by conventional topical authority.
What to do first: Audit whether AI engines can even parse your core pages and product data, fix your schema and feeds, then measure citation share across the top answer engines on a 30/60/90 day cadence rather than staring at classic rankings.

Why AI Search Visibility Is a Different Discipline From SEO

Classic SEO optimized for a ranked list of ten links where the human did the final choosing. AI search visibility optimizes for a synthesized answer where the machine does the choosing and often names only one to three sources. The stakes compress. In a ten-link world, being fourth still earned clicks. In an answer engine world, being the fourth-best match frequently means being omitted entirely, because the model only needed three citations to satisfy the query.

The retrieval mechanics differ too. Answer engines pull from a mix of their training data, live web retrieval, and increasingly from structured feeds and agent-facing protocols. That means your visibility depends on being present and legible in several layers at once: the open web content the model was trained on, the pages it retrieves in real time, and the machine-readable data it uses when a shopping or research agent needs to take an action. We have watched brands with excellent traditional SEO get skipped in AI answers simply because their structured data was thin or contradictory, and the model could not confidently extract a price, a spec, or an availability status.

Retrieval over ranking: Answer engines reward content that is easy to lift out cleanly. A page that buries the answer in the eighth paragraph loses to a page that states it in the first two sentences with supporting structure. We aim for the answer to be extractable within the first 60 to 80 words of any target section.

Entity over keyword: Models reason about entities and their relationships, not just keyword strings. If your brand, your products, and your category are not clearly defined and consistently described across your site and the wider web, the model has nothing stable to attach a recommendation to.

Trust over volume: In a synthesized answer, a single confident citation from a well-structured, consistent source beats ten thin pages. Answer engines are conservative about what they will state as fact, and they favor sources whose data does not contradict itself.

For a deeper foundation on how machine buyers change the retrieval landscape, our breakdown of the shift from predictive to agentic AI in ecommerce is a useful companion, because AI search visibility and agentic commerce are converging fast.

Checklist for framing the discipline correctly:

Reframe the goal: Track citations and mentions in AI answers, not just position on a link-based results page.
Prioritize extractability: State the answer early and support it with structure the model can lift cleanly.
Define your entities: Make brand, product, and category identities unambiguous and consistent everywhere.
Cover all retrieval layers: Optimize trained content, live-retrievable pages, and machine-readable feeds together.
Value consistency: Eliminate contradictions in price, spec, and availability data across every surface.

Getting Started: Auditing Whether AI Engines Can Even See You

Before optimizing anything, we run a visibility audit, because most teams have never actually checked what the machines perceive. The audit answers one blunt question: when a relevant query is asked, can leading answer engines find, parse, and confidently cite our pages and data? Roughly half the stores we audit fail on parsing alone, not on content quality.

Start with live prompt testing. Take your 20 most commercially important queries, the ones a real buyer or researcher would type, and run them through ChatGPT with browsing, Perplexity, Gemini, and Google AI Overviews. Record three things for each: whether you appear at all, whether the information cited about you is accurate, and which competitors are named instead. This gives you a baseline citation share, which is the single most important AI search visibility metric we track.

Next, check machine legibility. Fetch your key pages the way a crawler would, with JavaScript rendering disabled, and confirm the core facts, price, product name, key specs, availability, are present in the raw HTML and in structured data, not injected client-side after load. A shocking number of modern storefronts render critical product data only in the browser, which means a lightweight retrieval pass sees an empty shell.

Then inspect your structured data coverage. Validate that Product, Offer, Organization, FAQ, and Article schema are present and error-free on the page types that matter. Contradictions here are fatal: if your schema says the price is 49 dollars and your visible page says 59, the model may distrust both and skip you.

What this achieves: A baseline audit tells you whether your problem is invisibility (the model cannot see you), inaccuracy (it sees you but gets facts wrong), or preference (it sees you correctly but chooses someone else), and each of those requires a completely different fix.

Checklist to complete your starting audit:

Baseline citation share: Test your top 20 queries across four answer engines and log appearance, accuracy, and competitors.
Raw HTML check: Confirm core facts render server-side, not only after client JavaScript executes.
Schema validation: Verify Product, Offer, Organization, FAQ, and Article schema are present and error-free.
Contradiction scan: Ensure price, availability, and specs match across schema, visible page, and feeds.
Diagnosis label: Classify each gap as invisibility, inaccuracy, or preference before choosing a fix.

Core Setup: Structured Data and Machine-Readable Product Feeds

The foundation of durable AI search visibility is data that machines can consume without guessing. This splits into two workstreams: on-page structured data (schema markup) and off-page machine-readable feeds and protocols that agents can query directly.

On the schema side, we implement JSON-LD for every high-value page type. For products, that means complete Product and Offer objects with price, currency, availability, GTIN or SKU, brand, and shipping and return details. For informational content, Article and FAQPage schema help answer engines extract question-answer pairs directly, which is exactly the format they synthesize into responses. For your business itself, Organization and, where relevant, LocalBusiness schema establish the entity so the model knows who you are and can connect all your other content to a stable identity.

The feed side is where AI search visibility increasingly overlaps with agentic commerce. When an AI shopping assistant recommends a product, the next step is frequently an action: comparison, cart, checkout. That requires structured, real-time product data the agent can trust. Traditional product feeds built for Google Shopping are a starting point, but they were designed for ad platforms, not autonomous agents. Our guide to product feed optimization for AI shopping agents covers the distribution mechanics in depth, and AI shopping product feeds and UCP explains why a protocol layer beats a pile of one-off feeds.

Here is the market context that matters. According to UCP Checker, which independently monitors 18,124+ storefronts, roughly 72 percent pass full UCP validation, which works out to 13,007 verified stores. That figure reflects the stores UCP Checker tracks, a set that skews heavily toward Shopify, so it is not a claim about ecommerce at large. It is also worth keeping honest about what a passing manifest means: a conformant UCP manifest is not the same as an agent being able to complete a real checkout. Structured data gets you seen; execution reliability gets you chosen and transacted.

Real-time accuracy: Answer engines and shopping agents penalize stale data hard. If your feed says in stock and the item is not, the agent’s action fails, and repeated failures degrade the store’s trust in future retrieval. Our piece on why instant data sync matters for AI agents explains the mechanics of keeping data fresh enough to be actionable.

What this achieves: Complete, consistent, real-time structured data turns your catalog and content into something an AI can quote and act on with confidence, which is the precondition for every downstream visibility gain.

Checklist for your core data setup:

JSON-LD everywhere: Deploy Product, Offer, Organization, FAQPage, and Article schema across relevant templates.
Complete offer data: Include price, currency, availability, GTIN or SKU, brand, shipping, and returns.
FAQ extraction blocks: Structure question-answer pairs so engines can lift them directly into answers.
Agent-ready feeds: Move beyond ad-platform feeds toward a protocol layer agents can query and act on.
Real-time sync: Keep availability and pricing accurate within minutes, not hours, to preserve agent trust.

The Implementation Steps: A Sequenced Rollout

Teams stall when they try to do everything at once. We sequence AI search visibility work in a specific order so each step compounds on the last. Here is the rollout we use with clients, as an ordered plan you can run over roughly six to eight weeks.

First, fix legibility. In week one, resolve any server-side rendering gaps and validate schema on your top page templates. Nothing else matters if the machine cannot read the page. Aim for zero critical schema errors and all core facts present in raw HTML.

Second, establish entity clarity. In weeks two and three, tighten your Organization schema, ensure your brand name, description, and category are identical across your site, your social profiles, and any directories or marketplaces. Publish a clear, factual “about” and category page that a model can use as a canonical description of who you are and what you sell.

Third, build answer-shaped content. In weeks three and four, take your top 20 buyer questions and create or rewrite content so each answer is stated cleanly in the first two sentences, backed by FAQPage schema and supporting detail. This is the content most likely to be lifted verbatim into an answer.

Fourth, upgrade your feeds and connect an agent-ready protocol. In weeks four through six, move from ad-oriented product feeds to a real-time, agent-consumable data layer. For Shopify merchants specifically, our guides on Shopify AI automation and the agentic plan and selling in AI channels via the Shopify agentic plan map the platform-specific path.

Fifth, measure and iterate. From week six onward, re-run your citation share tests weekly and treat gaps as a backlog. What this achieves: A sequenced rollout means you never waste content effort on pages the machine cannot read, and you never chase feed optimization while your entity is still ambiguous.

Checklist for the rollout sequence:

Week 1 legibility: Zero critical schema errors, all core facts server-rendered.
Weeks 2 to 3 entity: Consistent brand identity and canonical category description everywhere.
Weeks 3 to 4 content: Answer-shaped pages for your top 20 buyer questions with FAQ schema.
Weeks 4 to 6 feeds: Real-time, agent-ready product data layer live and validated.
Week 6+ iteration: Weekly citation share testing feeding a prioritized fix backlog.

The RECALL Framework for Durable AI Search Visibility

We use a named framework so teams remember the pillars under pressure. RECALL stands for Retrievability, Entity clarity, Consistency, Answer shaping, Live data, and Links of trust. Each step has a job.

Retrievability: Make every important fact machine-readable in raw HTML and structured data. What this achieves: It guarantees the model can actually extract you, which is the non-negotiable base of visibility, and without it every other effort is invisible.

Entity clarity: Define your brand, products, and categories unambiguously and consistently across the web. What this achieves: It gives the model a stable identity to attach recommendations to, so citations accumulate to you rather than scattering across near-duplicates or getting attributed to a competitor.

Consistency: Eliminate contradictions between your schema, your visible content, your feeds, and third-party mentions. What this achieves: It raises the model’s confidence to the point where it will state facts about you rather than hedging or skipping you, because conflicting data is the fastest way to get filtered out.

Answer shaping: Structure content so the answer is stated cleanly and early, in the format engines synthesize. What this achieves: It maximizes the chance your exact wording is lifted into the generated answer, which is the strongest possible form of AI search visibility.

Live data and Links of trust: Keep product and availability data real-time, and earn mentions from sources the model already trusts. What this achieves: It moves you from merely visible to actively preferred and transactable, because agents favor sources that are both fresh and corroborated by reputable third parties.

In AI search, you do not rank, you get retrieved, and the store whose data is cleanest and most consistent is the store the machine trusts enough to recommend.

Checklist to apply RECALL:

Retrievability score: Every core fact present in raw HTML and schema, verified with rendering off.
Entity file: A canonical brand and category description used consistently across all surfaces.
Consistency audit: No contradictions between schema, page, feed, and third-party data.
Answer-first content: Target answers stated in the first two sentences of each section.
Freshness and corroboration: Real-time data plus mentions from trusted external sources.

Turn AI Search Visibility Into Agentic Revenue With UCPhub

Getting cited in an AI answer is only valuable if the agent can then act, comparing, adding to cart, and checking out without friction. That is exactly the gap the Universal Commerce Protocol closes, giving AI agents a standardized, real-time way to read your catalog and complete transactions instead of stopping at a recommendation they cannot execute. UCPhub connects your store to the agentic channels where discovery increasingly happens, so your improved AI search visibility converts into actual orders rather than lost intent. If you want to see whether your store is agent-ready and where the gaps are, talk to our team at UCPhub and we will map the fastest path from visible to transactable.

For the strategic context on why a protocol beats stitching together bespoke connectors, read our comparison of UCP versus custom AI integrations, which explains why point solutions do not scale as agent traffic grows.

Optimization: Increasing Citation Share Over Time

Once the foundation is in place, optimization is about steadily raising the share of relevant answers that name you. We treat this like a ranking problem with different signals, and we push on three fronts at once.

Content depth on the entities that matter: Answer engines prefer sources that demonstrate topical authority, not just a single thin page. If you sell running shoes, having genuinely useful, well-structured content on gait, cushioning, and use cases makes the model more confident citing you for shoe queries. Aim for coverage that answers the follow-up questions, not just the head query, because answer engines chain reasoning and reward sources that satisfy the whole thread.

Third-party corroboration: Models cross-check. When independent, reputable sources describe your product the same way your own pages do, confidence rises. Earned mentions, accurate marketplace listings, and consistent directory data all feed this. We prioritize fixing incorrect third-party data before chasing new mentions, because one wrong price on a big aggregator can suppress you everywhere.

Freshness signals: Update cadence matters more in AI search than most teams expect. Stale content and stale feeds both hurt. We set a minimum refresh cadence, availability within minutes, editorial content reviewed at least quarterly, so the model always sees current data.

For agencies managing many stores, the maintenance math changes dramatically once you move to a protocol layer. Our case-driven breakdown of how agencies went from ten feeds to one protocol and cut maintenance eighty percent shows why consolidation is both a cost and a visibility win, since fewer feeds means fewer contradictions for the model to trip over.

What this achieves: Optimization moves you from occasionally cited to reliably preferred, because you compound authority, corroboration, and freshness rather than relying on a single well-optimized page.

Checklist for ongoing optimization:

Topical clusters: Build depth around your core entities, not isolated pages.
Follow-up coverage: Answer the chain of related questions, not just the head query.
Third-party accuracy: Fix wrong external data before pursuing new mentions.
Refresh cadence: Availability updated within minutes, editorial reviewed quarterly.
Feed consolidation: Reduce feed count to cut contradictions and maintenance load.

Common Mistakes to Avoid

We see the same failures repeatedly, and most are quiet, the kind that cost visibility for months before anyone connects the dots.

Rendering the answer client-side only: If your price, specs, or availability appear only after JavaScript runs, a lightweight retrieval pass sees nothing. Fix: server-render or statically embed all core facts.

Contradictory data across surfaces: A price in schema that disagrees with the visible page, or a feed that disagrees with both, teaches the model to distrust you. Fix: single source of truth feeding every surface.

Chasing keywords instead of entities and answers: Stuffing keyword variants does nothing when the model reasons over entities. Fix: define entities cleanly and shape content as direct answers.

Treating a passing manifest as done: A conformant UCP manifest is meaningful, but as we noted, it is not the same as an agent being able to complete a real checkout. Fix: test the full path, discovery through transaction, not just validation status. Our UCP Hub demo testing guide walks through how to verify end-to-end.

Ignoring the difference between protocols: ACP and UCP are not interchangeable, and betting on the wrong integration approach wastes effort. Our ACP versus UCP strategic guide and the broader guide to agentic AI protocols including MCP, A2A, and UCP help you choose deliberately.

Checklist of mistakes to eliminate:

Client-side-only facts: Server-render every core detail.
Data contradictions: Enforce a single source of truth.
Keyword stuffing: Optimize entities and answers instead.
Validation complacency: Test discovery through checkout, not just the manifest.
Protocol confusion: Choose your integration approach deliberately, not by default.

Advanced Tips for Teams Already Ranking in AI Answers

Once you are reliably cited, the frontier work is about becoming the default recommendation and being transactable inside the agent experience.

Optimize for the transaction, not just the mention: The next competitive edge is agents that can complete a purchase without leaving the conversation. Building toward a UCP-first commerce model where AI agents are the primary shoppers means your visibility converts to revenue directly. This is where AI search visibility stops being a marketing metric and becomes a sales channel.

Instrument agent traffic: Most analytics do not distinguish agent-driven sessions from human ones. We tag and monitor agent behavior separately so we can see which answer engines drive real actions, not just impressions.

Pre-answer the comparison: Agents frequently generate comparison tables. Structure your product data so a fair comparison favors you on the attributes you genuinely win, and make those attributes unambiguous in your schema.

Build the strategic playbook: For a full commercial view, our strategic guide to agentic commerce in 2026 and the merchant playbook for AI shopping assistant integration tie visibility to revenue outcomes.

Checklist for advanced teams:

Transaction readiness: Enable in-conversation purchase completion via protocol.
Agent analytics: Segment and measure agent-driven sessions distinctly.
Comparison optimization: Structure data to win fair, machine-generated comparisons.
Default positioning: Aim to be the recommended option, not merely a listed one.
Revenue attribution: Tie citation share to actual agentic orders.

Measuring Success: 30/60/90 Day KPIs

We measure AI search visibility against outcomes, not activity. Here is the cadence we hold clients to, formatted so you can lift it straight into a dashboard.

By 30 days:

Legibility resolved: Zero critical schema errors and 100 percent of core facts server-rendered on priority templates.
Baseline citation share captured: Documented appearance rate across ChatGPT, Perplexity, Gemini, and Google AI Overviews for your top 20 queries.
Accuracy corrected: Every factual error in AI answers about your brand identified and traced to its source.

By 60 days:

Citation share lift: A measurable increase, we target 15 to 30 percent relative improvement, in the share of target queries where you are named.
Feed readiness: Real-time, agent-consumable product data live and validated end to end, not just at the manifest level.
Third-party consistency: Major external listings corrected so no contradictory price or availability data remains.

By 90 days:

Preference position: You appear in the top one to three cited sources for the majority of your priority queries.
Transactable path: Agents can move from recommendation to a completable action for your core catalog.
Attributed revenue: Agent-driven sessions instrumented and generating measurable orders, closing the loop from visibility to sales.

Checklist for measurement discipline:

Weekly citation tests: Re-run the top-20 query panel every week, not monthly.
Accuracy log: Track and resolve every factual error the engines state about you.
Freshness audit: Confirm data sync latency stays within target thresholds.
Revenue tie-out: Attribute agentic orders back to visibility improvements.
Backlog discipline: Convert every gap into a prioritized, owned fix.

If you are just getting started, do not touch content or feeds yet: fix legibility and entity clarity first, because a machine that cannot read or identify you will ignore even your best content. If you are auditing something that already exists and mostly ranks, skip ahead to citation share testing and third-party accuracy, since your gains there are usually about consistency and corroboration rather than net-new pages. Either way, resist the urge to optimize everything at once, because the sequence is what makes the effort compound.

Next Steps:

Run the top-20 query panel across all four answer engines today and record your baseline citation share.
Fetch your five highest-value pages with JavaScript disabled and confirm every core fact is present in the raw HTML and schema.
Book a readiness review with our team to map the path from visible to transactable, starting at ucphub.ai/contact.

Frequently Asked Questions

How do I improve visibility in AI search engines?

Start by making sure the machine can actually read you, because most visibility problems are legibility problems in disguise. Fetch your key pages with JavaScript rendering disabled and confirm your core facts, product names, prices, availability, and specs, appear in the raw HTML and in valid structured data. If those facts only render client-side, a retrieval pass sees an empty shell and you become invisible regardless of content quality.

Once legibility is solid, focus on entity clarity and answer shaping. Define your brand, products, and categories consistently everywhere, and rewrite your most important pages so the answer to the buyer’s question is stated cleanly in the first one or two sentences, backed by FAQPage schema. Answer engines lift clean, early, well-structured statements directly into their responses, so this format gives you the best chance of being quoted.

Then measure and iterate on citation share. Run your top commercial queries through ChatGPT, Perplexity, Gemini, and Google AI Overviews on a weekly cadence, record where you appear and where competitors appear instead, and treat every gap as a fixable backlog item. Improvement in AI search visibility comes from this loop of test, diagnose, fix, and re-test, not from a one-time optimization pass.

What improves visibility in answer engines?

Three levers move the needle most. First, machine-readable structured data, complete and error-free Product, Offer, Organization, FAQPage, and Article schema, gives the engine facts it can extract and cite with confidence. Second, consistency across every surface, so your schema, visible page, product feeds, and third-party listings never contradict each other, because contradictions cause the model to distrust and skip you. Third, real-time data accuracy, since answer engines and shopping agents penalize stale availability and pricing hard.

Topical authority and corroboration reinforce those levers. Building genuine depth around your core entities, and earning accurate mentions from reputable third-party sources, raises the model’s confidence to name you rather than a competitor. Answer engines cross-check, so when independent sources describe your products the same way you do, your citation likelihood climbs.

Finally, freshness and extractability tie it together. Content and feeds updated on a tight cadence, with answers structured to be lifted cleanly, consistently outperform larger but staler or messier sources. The winning pattern is clean, consistent, current, and corroborated, not simply large.

How do I get featured in AI-powered search results?

Getting featured, meaning cited by name inside a generated answer, requires being the easiest correct source for the model to use. That means your target answer is stated early and unambiguously, wrapped in valid FAQPage or Product schema, and free of any contradiction with your other data. When the model assembles a response, it reaches for the source it can extract cleanly and trust fully, and that is the position you are engineering toward.

Beyond the page level, being featured depends on entity strength and corroboration. If your brand identity is consistent across your site and the wider web, citations accumulate to a single stable entity rather than scattering. If reputable third parties confirm your facts, the model’s confidence rises to the threshold where it will state them outright rather than hedging or omitting you.

For commerce specifically, the ultimate form of being featured is being transactable inside the agent experience, not just mentioned. When an AI shopping agent can move from recommending you to completing a purchase through a protocol like UCP, your visibility converts directly to revenue. That is the frontier we help merchants reach, because a citation that cannot be acted on leaves money on the table.

Is AI search visibility different from traditional SEO?

Yes, meaningfully, though they share a foundation. Traditional SEO optimizes for position in a ranked list of links where the human makes the final choice, so being fourth still earns clicks. AI search visibility optimizes for being retrieved and cited inside a synthesized answer that often names only one to three sources, so being the fourth-best match frequently means being omitted entirely. The stakes compress and the tolerance for ambiguity drops.

The signals overlap but weight differently. Structured data, entity clarity, and data consistency matter far more in AI search than in classic SEO, while some traditional tactics like link volume matter less than trust and corroboration. Extractability, stating the answer cleanly and early, becomes a first-class ranking factor rather than a nice-to-have.

The practical takeaway is that you should not abandon SEO fundamentals, since topical authority and crawlability still help, but you must add the AI-specific disciplines of machine legibility, entity definition, real-time accuracy, and citation share measurement. Teams that treat AI search as just more SEO tend to plateau, while teams that treat it as its own discipline keep gaining.

How do structured data and product feeds affect AI search visibility?

Structured data is the vocabulary answer engines use to understand your pages without guessing. Complete, valid schema tells the model exactly what your product is, what it costs, whether it is available, and who makes it, so it can extract and cite those facts confidently. Thin or contradictory schema forces the model to infer, and models tend to skip sources they cannot parse cleanly, which quietly erases visibility even for high-quality pages.

Product feeds extend this into the agentic layer. When an AI shopping assistant moves from recommending to acting, it needs structured, real-time product data it can trust. Ad-platform feeds were built for a different job, so a protocol layer designed for agents, keeping availability and pricing accurate within minutes, is what lets your visibility translate into completed actions rather than failed ones.

The honest caveat is that a conformant manifest is not the same as a completable checkout. According to UCP Checker, which independently monitors 18,124+ storefronts, roughly 72 percent pass full UCP validation, a set that skews heavily to Shopify, but passing validation and reliably transacting are different milestones. You want both the data legibility that gets you seen and the execution reliability that gets you chosen.

How do I measure whether my AI search visibility is improving?

The core metric is citation share, the percentage of your priority queries where an answer engine names you as a source. Build a fixed panel of your top 15 to 20 commercial queries, run them across ChatGPT, Perplexity, Gemini, and Google AI Overviews weekly, and record appearance rate, factual accuracy, and which competitors show up instead. Watching this trend week over week tells you far more than classic rank tracking does.

Layer in accuracy and freshness metrics. Track every factual error the engines state about your brand and resolve it at its source, and monitor your data sync latency so availability and pricing stay current within your target window. These are leading indicators, because clean, fresh, accurate data precedes citation gains.

Finally, close the loop to revenue. Instrument agent-driven sessions separately from human traffic so you can attribute completed actions and orders to your visibility work. On a 30/60/90 day cadence, we look for legibility resolved by day 30, a 15 to 30 percent relative citation share lift by day 60, and a top one-to-three cited position plus attributed agentic revenue by day 90.