Agentic commerce case study: how a US nutrition brand became the store AI agents buy from, and doubled D2C revenue
How we made a California nutrition and supplements brand the store an AI agent can search, trust, and check out from without a human, then won the bottom of the funnel and fixed the reputation AI reads, and doubled its direct-to-consumer revenue in six months.
Krishna Kaanth MFounder, MaximusLabs
Case study10 min read
The US supplement market is one of the most crowded corners of e-commerce, and the way people buy it is changing fast. The brand that wins the next decade is the one an AI shopping agent can search and check out from without a human in the loop, the one AI engines name in their answer, and the one shoppers trust after they have read the whole internet about it.
Our client had the product, the reviews, and a loyal base. What it did not have was visibility where the next wave of buyers, and the agents buying on their behalf, were actually looking. High-intent demand was leaking to competitors and, increasingly, to AI answers that never mentioned the brand at all.
A leading US nutrition and supplements brand
California based, selling direct to consumers across the United States, with a family of sub-brands spanning daily supplements, protein and greens, and an everyday foods line that includes granola.
Direct-to-consumerThird-party testedSubscription and one-timeMulti sub-brand catalog
2.0x
direct-to-consumer revenue, in six months and still climbing
Client analytics
+212%
sessions referred by AI search engines
Referral analytics
Top 20
bottom-of-funnel keywords won on Google and in AI answers
Rank tracking
38→91
agent-readiness score, start to month six
Readiness audit
Their goals were clear
Make the storefront the easiest in the category for an AI agent to search, verify, and check out from, end to end without a human.
Win the bottom of the funnel across Google and AI answers for the top 20 high-intent keywords in the category.
Become the brand AI engines recommend, instead of one they quietly route around.
Fix the sentiment AI reads across Reddit, forums, and review sites, where models decide who to trust.
Grow direct-to-consumer revenue, measurably, and tie every move back to it.
A discovery audit showed exactly where the demand was leaking before it ever reached the cart.
0
of six agent-commerce capabilities supported at the start: no callable search, feed, agent endpoint, or unaided checkout
Discovery audit
61%
of priority bottom-of-funnel queries returned an AI answer that never named the brand
AI answer audit
25%
of monitored brand conversation was negative or unanswered on the forums AI leans on
Sentiment baseline
When five brands sell the same granola, the one an agent can search, verify, and check out from wins the order. Product parity plus a navigability advantage equals the sale.
The thesis we set out to prove
1
Keyword research, the way buyers actually search
Most e-commerce keyword research is done backwards. It chases volume instead of intent, and it ignores the five stages of customer awareness that decide which words a buyer types at each step toward a purchase.
We started at the most-aware end, where the credit card is already out, then worked up the funnel. For every product line we built the keyword set from three sources: the queries the brand already half-ranked for, the gaps where competitors ranked and the brand did not, and the unscripted language real buyers use on Reddit, Quora, and YouTube. That last source is where the highest-converting long-tail phrasing hides, and almost no competitor mines it.
Exhibit 1
The five stages of buyer awareness, and the content we built for each.
The question the buyer asks at each stage, the asset we published to answer it, and the surface it was engineered for.
Stage 1Unaware
Why am I so tired every afternoon?
Plain-language guides on energy, sleep, and nutrient gaps. Asset: educational explainers.
AI Overviews, Perplexity
Stage 2Problem aware
Do magnesium or electrolytes actually help?
Ingredient science and symptom-to-solution content. Asset: problem guides.
ChatGPT, Gemini
Stage 3Solution aware
Which form of magnesium is best for sleep?
Comparison and how-to-choose content. Asset: solution guides.
Agent-ready product pages, subscriptions, post-purchase. Asset: the checkout.
Agents, email
Source: MaximusLabs buyer-journey map and content plan for the engagement, based on the five stages of customer awareness.
How we found the keywords
Three techniques, run in order: tap the brand's existing query data for terms it already half-ranked for, reverse-engineer competitor keyword gaps for high-intent terms with reachable difficulty, then mine Reddit, Quora, and YouTube for the unscripted phrases buyers use that no keyword tool surfaces.
2
Winning the bottom of the funnel, in search and in AI
We optimized relentlessly for the queries closest to purchase, then tracked something most agencies still ignore: not just the Google rank, but whether AI engines cited the brand in their answer for that query.
Exhibit 2
Bottom-of-funnel keyword performance, ranked by the metric that matters most.
Switch the metric to re-rank. AI answer share is how often the brand is cited in AI answers for the query. Organic CTR is click share from the search result.
best sugar free electrolyte powderhigh intent · now position 1
74%AI answer share
third-party tested creatinehigh intent · now position 1
71%AI answer share
best magnesium glycinate for sleephigh intent · now position 2
66%AI answer share
best greens powder redditsocial intent · now position 2
58%AI answer share
protein granola for weight losshigh intent · now position 3
62%AI answer share
is the brand third-party testedbrand intent · now position 1
69%AI answer share
vegan protein powder for womenhigh intent · now position 4
54%AI answer share
collagen peptides for joint painhigh intent · now position 3
60%AI answer share
Source: Rank and AI-citation tracking across Google, ChatGPT, Perplexity, and Gemini, month six. Eight of the top twenty shown.
Bottom-of-funnel keyword
Intent
Before
After
In AI answers
best sugar free electrolyte powder
Purchase
22
1
Cited
third-party tested creatine monohydrate
Purchase
Not ranking
1
Cited
best magnesium glycinate for sleep
Purchase
19
2
Cited
protein granola for weight loss
Purchase
14
3
Cited
is the brand third-party tested
Brand
6
1
Cited
best greens powder reddit
Social
Not ranking
2
Cited
collagen peptides for joint pain
Purchase
16
3
Cited
The technical and on-page foundation that makes it readable
Rankings do not hold without a clean technical base, and the same fixes that satisfy Google now also decide whether an AI crawler and a shopping agent can read the page at all. We rebuilt the foundation around five disciplines.
1
Crawl depth and internal links
Every priority page brought within three clicks of the homepage, with internal links routing authority to the bottom-of-funnel pages that convert.
2
Schema and structured data
Product, Offer, FAQ, and Article schema across the catalog, so search engines and AI get price, stock, and testing data as machine-readable facts, not prose to guess at.
3
Clean, crawlable HTML
Critical content rendered in HTML rather than locked behind JavaScript, so AI crawlers and agents read the product without executing scripts.
4
Core Web Vitals and image weight
Compressed, correctly sized WebP imagery and sub-three-second loads, because 40% of shoppers abandon a slow store before it renders.
5
On-page built for NLP, not keyword density
Answer-first 40 to 80 word nuggets and full entity coverage of each topic, so the page reads as the most relevant, citable answer rather than a stuffed one.
3
Online reputation management, because AI reads the whole web
AI answers do not cite your storefront alone. They synthesize the entire internet about you, then decide whether to recommend you. So we analyzed sentiment across the sources models lean on, then ran precise, honest interventions to move it.
We do not buy fake praise. We find the real, unanswered questions and the genuine objections in the places models read, then answer them in the open, accurately and helpfully, on Reddit, Quora, YouTube, professional networks, and review sites. We also helped the brand collect first-party reviews on the platforms buyers and AI trust most. Sentiment follows substance, and AI follows sentiment.
Exhibit 3
Brand sentiment across monitored sources, before and after.
Share of monitored mentions by sentiment. Hover a slice or a legend row for detail.
Before the engagement
Positive41%
Neutral34%
Negative25%
After six months
Positive73%
Neutral19%
Negative8%
Source: MaximusLabs sentiment monitoring across Reddit, Quora, YouTube, professional networks, and review sites.
Exhibit 4
Positive share of voice by platform, after the intervention.
Share of brand conversation that is positive, by source. The before figure is shown beneath each label.
Redditwas 39%
68% +29
Quorawas 45%
72% +27
YouTubewas 52%
74% +22
Professional networkswas 48%
70% +22
Review siteswas 57%
81% +24
Source: MaximusLabs reputation tracking, before versus month six. Interventions were honest, sourced, and disclosure-compliant.
The MaximusLabs ViewKrishna Kaanth M, Founder
Online reputation is no longer a PR line item. It is a ranking input for every AI answer that decides whether a buyer ever hears your name.
If the conversation about you is thin or negative where models read, you get left out of the answer, no matter how good your page is. We fix the substance first, and the sentiment follows.
4
Engineering the storefront for agentic commerce
This is the part almost no competitor has touched yet, and the one we believe decides the next decade of e-commerce. We researched the emerging rails, the Universal Commerce Protocol, Agent-to-Agent, and WebMCP, then implemented them in layers so an AI agent can find a product and buy it without a human.
Exhibit 5
Why navigability decides the order when products are at parity.
Each bubble is a brand in the category. Hover for detail. Bubble size reflects relative share of agent-initiated orders.
Product quality →
High quality · Low navigabilityInvisible to agents
High quality · High navigabilityAgent favorite
Low quality · Low navigabilityOut of the race
Low quality · High navigabilityNavigable but weak
Beforethe client
Afterthe client
Comp A
Comp B
Agent navigability →
Source: MaximusLabs category readiness assessment. Positions scored on agent navigability and independent product-quality signals.
Exhibit 6
Agent-readiness score, start to month six.
A composite of crawlability, structured product data, callable search, action surface, and trust signals. Out of 100.
91
out of 100
A near-bottom store became a category leader for agents
The same catalog and the same products. What changed is that an agent can now discover, read, and transact against the store without a human in the loop.
38
at the start
→
91
month six
Source: MaximusLabs Agent-Readiness Audit, engagement start versus month six. Composite of crawlability, structured data, callable search, action surface, and trust signals.
Exhibit 7
What an AI agent hit at the start, versus what it can do today.
The capability stack we researched and shipped: callable search, structured data, a feed, unaided checkout, WebMCP, and Agent-to-Agent plus Universal Commerce Protocol rails.
×At the start, the agent hit a wall
No on-site product search an agent could call
Product data trapped in images and scripts, not machine-readable
No product and offer feed for price and stock
Checkout required a human session and a captcha
No WebMCP tool surface to act against
No Agent-to-Agent or Universal Commerce Protocol support
✓Today, the agent completes the job
Calls a clean structured search and gets the exact product
Reads full product, price, and stock data on every page
Pulls a machine-readable product and offer feed
Completes checkout end to end without a human
Acts through a published WebMCP tool surface
Transacts over Agent-to-Agent and Universal Commerce Protocol rails
Source: Implementation log, MaximusLabs agentic commerce workstream. Protocols per current public specifications.
Exhibit 8
Can the agent get from landing to a confirmed order? Before, no. Now, yes.
The same six-step purchase path, run as an AI agent. Hover any step for what happened.
Before
Land
Reached
Search
Blocked
Match
Failed
Add to cart
Failed
Checkout
Failed
Confirm
Failed
After
Land
Done
Search
Done
Match
Done
Add to cart
Done
Checkout
Done
Confirm
Purchase
Source: Agent task replay, MaximusLabs. Same purchase path executed by an autonomous agent at the start versus month six.
Exhibit 9
Conversion lift on the agent-ready storefront, by category.
Order conversion for agent-initiated sessions, standard storefront versus the agent-ready build. Hover a bar for detail.
2.1%
7.6%
2.4%
8.3%
1.9%
7.1%
2.6%
8.8%
Granola3.6x lift
Protein3.5x lift
Magnesium3.7x lift
Electrolytes3.4x lift
Standard storefront
Agent-ready build
Source: Agent-initiated session conversion, MaximusLabs instrumentation. Median lift 3.6x across categories.
Why this is the first level, and not the last
An on-site search an agent can call is the entry ticket, because most stores still do not have one. The durable moat is everything behind it: crawlability, structured data, a WebMCP tool surface, and Agent-to-Agent plus Universal Commerce Protocol support, so agents transact directly rather than scraping their way through.
5
Measuring what compounds
One integrated six-month program across four workstreams, instrumented from day one so every move tied back to revenue. We tracked Google rank, AI citation share across every major engine, sentiment, and agent-readiness, weekly. If the brand was not in the answer, the work was not done.
Exhibit 10
The six-month engagement, workstream by workstream.
Hover a bar for its window. The navy milestone marks when revenue crossed 2x.
Source: MaximusLabs engagement plan. Continuous instrumentation across all four workstreams.
Model the agent channel for your own catalog
Move the sliders to your numbers. The model applies the median conversion lift we observed when an agent can complete checkout without a human, capped at a realistic ceiling.
MaximusLabs Model
The agent channel opportunity
Monthly storefront sessions, including agents40,000
Current conversion rate2.4%
Average order value$58
Incremental revenue, per year
$0
Orders today, per month0
Orders agent-ready, per month0
Added orders, per month0
Added revenue, per month$0
Model applies a 3.6x median conversion lift for agent-completable checkout, capped at a 95% ceiling. This is a planning model, not a guarantee. Your result depends on catalog, category, and agent traffic mix.
The Results
Six months in, direct-to-consumer revenue had doubled, and AI was the fastest growing channel.
Revenue did not just grow, it changed shape. A rising share now arrives from AI search engines and agents that did not send a single buyer before the engagement.
Exhibit 11
Revenue doubled in six months, and the AI channel grew fastest of all.
Indexed to 100 at the start of the engagement. Solid line is observed, dashed line is the current run rate carried forward.
Observed
Run rate carried forward
Source: Client e-commerce and analytics platforms, engagement months 0 to 6, run rate carried to month 8. Revenue indexed to a baseline of 100 at the start.
The outcomes, against the goals they set
Direct-to-consumer revenue2.0x +100%
AI-referred sessions+212%
Priority BOFU keywords wonTop 20
New revenue from AI and agents38%
Positive brand sentiment41% → 73%
Agent-readiness score38 → 91
Agent-session conversion lift3.6x
Engagement window6 months
The numbers that frame the engagement
2.0x
direct-to-consumer revenue, in six months and still climbing
Client analytics
+212%
sessions referred by AI search engines
Referral analytics
20
priority bottom-of-funnel keywords won on page one and cited in AI answers
Rank tracking
91
agent-readiness score out of 100, up from 38 at the start
Readiness audit
The MaximusLabs ViewKrishna Kaanth M, Founder
In agentic commerce, distribution is no longer just where you show up. It is whether a machine can read you, trust you, and buy from you without help.
We did not change the product. We made the brand legible and purchasable to the systems that now stand between the buyer and the checkout, then made sure those systems had every reason to recommend it. Doubling revenue was the result of closing that gap end to end.
We do not just do SEO. We make your store the one AI agents buy from.
We had a product we believed in and sales that had plateaued. Six months later our e-commerce revenue had doubled, and the channel growing fastest was one we were not even visible in before.
Founder and CEOA leading US nutrition and supplements brand
The outcome in one line
2x revenue. 6 months. One agent-ready storefront.
Bottom-of-funnel SEO, buyer-journey content, reputation for AI, and a store agents can actually buy from. One integrated program, tied to revenue.
More case studies
See how we win AI search across categories.
From D2C e-commerce to B2B SaaS, the same playbook: be the answer AI engines cite, and the one agents can act on.
