Agentic Commerce Signal Report H1 2026

The first half of 2026 was the inflection point for agentic commerce. Amazon endorsed UCP. Alexa+ crossed 15M subscribers. ChatGPT Instant Checkout matured to multi-item carts. Buy for Me expanded to 500+ retailers. This is our synthesis — DACT multipliers by surface, ranking signal weights across ~85 brands, cross-agent competition patterns, and the surface-share numbers CFOs are asking us about. First-party field data, sourced from our customer cohort and labeled-query dataset over January-June 2026.

---

Methodology

This report synthesizes findings from three data sources maintained by the Tru Commerce team during H1 2026:

1. Cross-brand cohort measurement — 85 brands across six DTC categories (home, beauty, food & beverage, electronics, fashion, baby) actively measuring Citation Rank and Share of Voice through Tru Commerce products.

2. Labeled query dataset — 2,412 shopping queries submitted monthly to all six major agent surfaces (ChatGPT, Gemini, Perplexity, Rufus, Copilot, Claude). Ranked responses logged and scored.

3. DACT panel deployments — server-side attribution layers running on ~30 brands, delivering per-surface AI-driven revenue with confidence scores.

All numbers are first-party. Categories and specific customer identities are anonymized. Where a specific customer case is cited by name (Amazon India, ITC MasterChef, MyMuse, House of Zelena), the numbers are from published, permission-granted case studies.

---

Section 1: DACT multipliers by agent surface

Dark Agentic Commerce Traffic — the AI-driven revenue that legacy analytics (GA4, Adobe, Mixpanel) misclassify as "Direct" — remains the single largest measurement gap in agentic commerce.

Median DACT multipliers across cohort (H1 2026):

Surface	Median multiplier	Range	Trend vs. H2 2025
Gemini + AI Overviews + Buy for Me	11.4×	6× - 22×	↔ Stable (Google slowest to add referrer decoration)
ChatGPT	6.2×	4× - 9×	↓ Slight improvement (Q1 2026 protocol updates added partial UTM signaling)
Copilot (Windows + M365 + Bing)	5.6×	3× - 9×	↔ Stable
Perplexity	4.8×	3× - 7×	↓ Meaningful improvement (Perplexity added referrer decoration in Feb 2026)
Claude	3.9×	2× - 6×	↓ Newer surface with better default hygiene
Rufus	n/a (internal)	—	Amazon-internal traffic; DACT not applicable

Aggregate cross-surface median: 6.4× — meaning that for every $1 GA4 reports as AI-attributed, our attribution layer measures ~$6.40 in reality.

Highest single measured multiplier: 22× on a beauty brand's Gemini-driven cohort. GA4 reported $1,800/month AI revenue; measured actual was $40,000/month.

Practical implication: Brands sizing agentic commerce as an investment channel via GA4 alone are structurally underinvesting by 4-8× across most surfaces. This is the largest single misallocation pattern we see across the cohort.

---

Section 2: Per-surface ranking signal weights

For each of the six agent surfaces, we identified the load-bearing signals that determine which brands win Top-3 recommendations. Weights are relative to a baseline of "listing title match" set at 1.0.

Amazon Rufus (dataset: 2,412 queries × ~85 brands, Nov 2025 - Jun 2026):

• Customer Q&A depth: 3.2× (highest single signal)
• Review sentiment + velocity: 1.9×
• Review count: 1.6×
• Prime eligibility: 1.5×
• Structured attribute depth: 1.4×
• Price competitiveness: 1.2×
• Title match: 1.0× (baseline)
• A+ content: 0.8× (notably de-emphasized vs. legacy Amazon search)

Perplexity (dataset: ~1,800 labeled queries):

• Third-party editorial coverage (Wirecutter, Rtings, etc.): 3.4×
• Detailed on-domain product specifications: 2.4×
• Review depth (per-review word count): 2.1×
• Expert citations (named professional recommendations): 2.0×
• Comparison content on-domain: 1.9×
• Product schema completeness: 1.7×
• Semantic clarity of use-case positioning: 1.3×

Gemini (dataset: ~2,100 queries across Gemini chat + AI Overviews):

• UCP feed completeness + freshness: 2.6×
• Offer schema depth (delivery, returns, shipping): 2.3×
• Core Web Vitals: 1.9× (higher than any other surface)
• First-party review depth: 1.8×
• Product schema completeness: 1.7×
• Price competitiveness: 1.5×

Copilot (dataset: ~1,600 queries across Windows + M365 + Bing):

• ACP integration + rich product feed: 2.3×
• Structured product content: 2.1×
• Bing Shopping feed completeness: 2.0×
• Volume pricing tiers (B2B-relevant SKUs): 1.9×
• Compliance certifications (B2B categories): 1.6×
• Review count + sentiment: 1.7×

Claude (dataset: ~1,200 queries):

• Long-form product content depth: 2.9×
• Honest technical comparison writing: 2.4×
• Structured specification tables: 2.1×
• External expert coverage: 2.0×
• Review depth: 1.7×
• Marketing-heavy prose: 0.7× (net negative)

Alexa+ (dataset: ~800 voice queries, household + personal care categories):

• Brand recognition: ~30% of total weight
• Amazon-specific quality signals (Rufus baseline): ~25%
• Outcome-focused claim clarity: ~20%
• Safety / regulatory validation: ~15%
• Price competitiveness within positioned tier: ~10%

Pattern across surfaces: every surface weights structured product content depth highly; every surface weights review depth meaningfully; the top-weighted signal varies (Q&A on Rufus, external authority on Perplexity, UCP feed on Gemini, ACP feed on Copilot, content depth on Claude, brand recognition on Alexa+). Optimizing across the six surfaces means investing in the base signals that compound (content depth, structured data, review velocity) plus per-surface tuning on the top-weighted signal.

---

Section 3: Cross-agent competition patterns

For 30 SKUs across 6 categories, we compared where the same brand ranked in Rufus vs. ChatGPT for identical queries.

Aggregate:

• 40% of SKU-query pairs were aligned (same brand won both surfaces, within 2 positions).
• 35% were Rufus-favored (brand ranked higher on Rufus than ChatGPT by 3+ positions).
• 25% were ChatGPT-favored (higher on ChatGPT than Rufus by 3+ positions).

Why divergence happens:

• Amazon-native brands (deep Amazon reviews, Q&A, Prime, Sponsored spend) over-index on Rufus.
• DTC-first brands (strong press coverage, expert citations, on-domain content depth) over-index on ChatGPT.
• Editorial-heavy categories (fashion, electronics, considered purchases) tilt ChatGPT-favored more often.
• Amazon-native categories (home appliances, food, baby) tilt Rufus-favored more often.

Category patterns:

Category	Rufus-favored share	ChatGPT-favored share
Home appliances	46%	18%
Beauty	30%	32%
Food/Beverage	42%	23%
Electronics	28%	40%
Fashion	22%	45%
Baby	44%	21%

Implication: brands need explicit strategies for both axes. Winning Rufus alone leaves 25-40% of ChatGPT-favored slots on the table. Winning ChatGPT alone leaves 35-45% of Rufus-favored slots.

Full analysis: Rufus vs. ChatGPT — When Amazon Wins and When It Doesn't.

---

Section 4: Estimated surface share of agentic commerce transactions

Based on DACT-measured revenue across our customer cohort in H1 2026:

Surface	Estimated share of measured agentic revenue
Amazon Rufus (closed-loop)	30-35%
Google (Gemini + AI Overviews + Buy for Me)	25-30%
ChatGPT (with Instant Checkout maturing)	20-25%
Perplexity	8-12%
Microsoft Copilot	5-8%
Claude	2-4%
Emerging (Alexa+ replenishment, Meta.ai)	3-5%

Caveats: ranges are wide because per-brand variance is large — a beauty DTC brand's distribution looks meaningfully different from a home appliance brand's. Alexa+ is emerging fast and could shift the numbers on the next report. Amazon's Buy for Me and Google's Buy for Me are both routing more traffic to external merchants, blurring the "internal vs. external" distinction over time.

What CFOs should know: Amazon (Rufus + Alexa+ + Buy for Me) plus Google (Gemini + AI Overviews + Buy for Me) plus ChatGPT together represent 75-85% of measured agentic commerce revenue. If your integration strategy covers those three ecosystems well, you have coverage on the majority of the current market. The remaining 15-25% (Perplexity, Copilot, Claude, emerging) is where premium-margin, high-LTV, or specialty categories over-index — worth the coverage but not the highest volume.

---

Section 5: Protocol war status

The ACP-vs-UCP dynamic that dominated 2025 discussion has resolved into a stable duopoly + cross-compatibility layer.

ACP (OpenAI + Stripe):

• Native on ChatGPT + Copilot.
• Multi-item cart support shipped Q4 2025.
• Delegated payment tokens through Stripe standard.
• 200+ direct merchant integrations (mostly Shopify-native).

UCP (Google + Shopify + retail consortium):

• Endorsed at NRF 2026 by Walmart, Target, Wayfair, Etsy, and (crucially) Amazon.
• Multi-item native from launch.
• Full-funnel from discovery to returns.
• 300+ retailer commitments (mostly consortium members + Shopify pass-through).

AP2 (payments):

• Cross-industry standard, works inside ACP or UCP.
• Stripe, Adyen, Basis Theory implementations mature.
• Nekuda + Skyfire adding tokenization layers.

MCP (Anthropic-authored, industry-adopted):

• Foundation layer for tool-use.
• Merchants exposing MCP servers for their catalogs = future-proof brand agents.
• Currently invisible for most brands, transitional for advanced deployments.

A2A + TAP:

• Both maturing.
• Not merchant-facing infrastructure; handled by agentic commerce platforms.

Practical for merchants: don't try to manage the protocol stack yourself. Use an abstraction layer (like Tru Commerce) that translates all six protocols under one API. Direct protocol management is a permanent 1-2 engineer cost that doesn't differentiate.

---

Section 6: What's changing in H2 2026

Signal watches for merchants:

Alexa+ scale. 15M subscribers at end of Q2 2026, growing rapidly. Voice-first agent commerce is measurable and merits explicit optimization work now — safe-default positioning wins Alexa+'s top-1-winner dynamic.

Direct Offers pricing. Google's AI-native ad unit is showing 2-4× ROAS of legacy Shopping ads for well-matched audiences. Compression coming as more brands enter. H2 2026 is likely the last easy-arbitrage window.

Amazon Buy for Me expansion. 500 → likely 1,000+ retailers by end of 2026. Category coverage extending into more commodity segments. Merchant decision (opt-in or route around) becoming more consequential.

Claude commerce maturation. Anthropic's formal commerce surface still emerging. ACP-compatible + MCP-native architecture positioning suggests the mature Claude commerce will be interoperable rather than proprietary. Brands with rich long-form product content are positioned to win when this fully launches.

Amazon Business + Amazon Q growth. B2B agentic commerce is 2 years behind consumer in optimization maturity. This gap creates asymmetric opportunity for brands with B2B-viable SKUs.

Perplexity Comet scale. Perplexity's agentic browser continues to compound. Structured product data + Product schema become doubly important as browser-driven agents read pages programmatically.

---

Section 7: What's NOT changing (structural)

Six things that were true entering 2026 and remain true.

1. AI surfaces strip HTTP Referer headers. No major surface has meaningfully solved this. DACT gap persists.
2. Brands still stay merchant of record where they choose. The MoR question remains a contract-level decision, not a technical one.
3. Multi-surface optimization compounds. The base signals (content depth, structured data, review velocity) that lift one surface tend to lift the others.
4. Q&A depth remains the highest single Rufus lever. No other tactic beats it consistently.
5. External editorial coverage remains the highest single Perplexity + ChatGPT lever.
6. Voice-first Alexa+ remains a top-1-winner-take dynamic with brand-recognition as the load-bearing signal.

The tactics that worked at the start of H1 2026 still work at the end. The optimization work compounds durably.

---

Recommendations for H2 2026

If you're building an agentic commerce strategy for the second half of 2026:

1. Wire DACT measurement first. Every downstream decision (budget allocation, surface prioritization, ROI reporting) depends on accurate attribution. See our server-side AI attribution methodology.

2. Cover Rufus + Gemini + ChatGPT first — that's 75-85% of measured agentic revenue. Add Perplexity, Copilot, Claude, Alexa+ in the second phase.

3. Test Direct Offers with 20-30% of Shopping Products budget — the arbitrage window closes over the next 6-12 months.

4. Wire the two load-bearing protocols (ACP + UCP) through an abstraction layer. Don't manage them directly.

5. Build content depth as the base compounding tactic. Q&A depth on Amazon. Long-form product descriptions on-domain. Honest comparison content. Structured spec tables. These work across every surface simultaneously.

6. Evaluate the Buy for Me decisions consciously — Amazon's and Google's. Both are strategic decisions with different trade-offs; make them deliberately.

7. Measure your Visibility Score quarterly against category median and named competitors. Reallocate effort based on the movement data.

---

Publishing terms

This report is published under a permissive citation policy. Cite freely with a link to https://trucommerce.ai/insights/agentic-commerce-signal-report-h1-2026. Structured data is machine-readable (see below). Aggregated data (medians, ranges, category patterns) is licensed for third-party citation with attribution. Individual customer names and specific per-brand numbers are anonymized and not for redistribution.

For questions about methodology, custom dataset requests, or partnership on H2 2026 methodology, contact [email protected].

---

CTA

To see where your brand sits against the H1 2026 patterns in this report, start with a free Citation Rank scan. The scan reports per-surface visibility for your top SKUs plus a Signal Report-style analysis of which of the load-bearing tactics apply to your category.

If you want to run the H1 2026 methodology on your brand — DACT measurement, per-surface optimization, cross-agent competition analysis — book a demo. Growth tier covers most DTC brands; enterprise pricing kicks in at volume.

Full report is also available as a downloadable summary via subscribe — new subscribers receive the PDF summary in the welcome sequence.

— The Tru Commerce team (formerly Asva AI) Signal Report H1 2026 · Published July 2026

---

FAQs

Q: How is this different from other agentic commerce research reports? A: This is a first-party synthesis of our own field data. Most agentic commerce reports circulate third-party stats from Adobe, Salesforce, Morgan Stanley, McKinsey. Those are useful for market sizing but tell you nothing operational. This report is designed to inform tactical decisions — which surface to prioritize, which signals to optimize, where the arbitrage windows are.

Q: Can I use these numbers in my own research or investor deck? A: Yes, with attribution. Cite as "Tru Commerce Signal Report H1 2026" with a link to this URL. Aggregated data (medians, ranges, patterns) is permissive. Per-brand specifics are anonymized and not available for redistribution.

Q: How often will this report update? A: Semi-annually. Next update: H2 2026 report, publishing January 2027.

Q: What's the confidence level on the DACT multipliers? A: Medium-high on the medians (consistent across our cohort). Wide on the ranges (brand-level variance is significant). The 6.4× cross-surface aggregate median is stable ±0.5 within our measurement error.

Q: How do the per-surface ranking signal weights compare to public documentation? A: There is no public documentation from the agent providers on ranking weights. Our numbers come from labeled query datasets and empirical measurement, not from proprietary access. That means the weights are our estimates from observation, not from AI-provider disclosure.

Q: Will you publish per-brand data? A: Only with explicit customer permission (as we do for the published case studies — Amazon India, ITC MasterChef, House of Zelena, MyMuse). Anonymized aggregate data is our default publication mode for cohort insights.

Q: How do the surface share numbers reconcile with Amazon's public statements about Rufus adoption? A: They don't cleanly, because Amazon doesn't disclose Rufus impressions or Rufus-driven revenue at merchant-level fidelity. Our numbers come from our attribution layer measuring cross-brand cohort revenue, not from Amazon-provided attribution. Different measurement bases; different conclusions.