Blog
Conversion intelligence
for growth teams.
Buyer psychology and notes on what stops people from buying, written for teams running serious ad budgets.
Your rankings held and your organic traffic fell. Here is what AI search actually took.
Pew found clicks roughly halve when an AI summary appears. Seer found the loss lands almost entirely on the searches an AI can finish. This is not a new discipline replacing SEO. It is the old foundation, with clarity, trust, and machine readability carrying the weight.
Your A/B test did not have enough traffic to tell you anything
A meta-analysis of 115 real tests found about 70% could not detect the size of win they were looking for. That is worse than running no test at all, because a blind result still feels like evidence.
eLLMo is short for Easy LLM Optimization. The spelling is for our daughters.
The name comes up on almost every first call. One half of it is the job we set out to do for small businesses. The other half is that our daughters love Elmo.
D2C citation building: what actually puts your brand in an AI answer
Muck Rack analyzed more than 25 million cited links across ChatGPT, Claude, and Gemini. Earned media accounted for 84% of AI citations. Paid content accounted for 0.3%. For a direct to consumer brand, that changes where the work happens, and what counts as it working.
Fashion week is debating AI in the studio. Shoppers already brought it to the store.
Designers at the Fall/Winter 2026 shows split into three camps over AI: creative partner, back office tool, or thing to refuse outright. All three are positions on how clothes get made. None of them is a position on how clothes now get found.
Paid search needed a third off to close. AI referrals needed a quarter.
In one store's month of orders, nine in ten paid search orders needed a discount to close, averaging a third off. Orders from AI assistants needed one two thirds of the time, at a quarter off. Same store, same catalog, same month.
A trillion dollars will be spent by agents. What makes one pick your brand?
A joint ICSC and McKinsey report puts US agentic commerce at $1 trillion in revenue by 2030. The brands that win it will be the ones a machine can read, verify, and defend to the person who asked.
Holiday 2026: one in five visitors will arrive from an AI chat
Salesforce expects 20% of holiday ecommerce traffic to come from AI chat agents this year. Cart abandonment sits at 82%, a third of shoppers are trading down, and free shipping just got 7% more expensive. Here is the readiness list.
Discovery has left your website. The data on where it went.
Between August 2025 and May 2026, product discovery on brand-owned properties fell 7% and traditional search fell 15%, while AI assistants and other new channels grew 38%. Your site is no longer where people find you. It is where they check you.
The information was on the page. The buyer never found it.
Warranty terms inside a description. A return policy nowhere near the buy button. A review count too quiet to notice. Nothing was missing. It was all just somewhere the buyer was not looking at the moment they needed it.
AI traffic went from converting 38% worse to 42% better in twelve months
Adobe measured an eighty point swing in how AI-referred shoppers convert, across more than a trillion visits to US retail sites. The traffic did not get better. The buyers arrived with better briefs.
A third of retail product pages cannot be read by AI. Yours might be one.
Adobe found roughly 34% of retail product pages inaccessible to AI systems, and a quarter of homepage and category content not optimized for language models. This is happening while AI traffic grows at triple digits and converts better than any other source.
What buyer simulations reveal that your analytics never will
Five findings from recent runs across different categories: a subscription nobody could price, a commitment with no visible exit, a mobile default nobody chose, reviews about the wrong product, and two bugs that never threw an error.
Why subscription attach rates collapse on the product page
A coffee brand showed the same upfront price for a one-time bag and a subscription. Test buyers chose one-time again and again, and told us exactly why: the page never did the math for them.
Cancel anytime is not a detail. It is conversion copy.
A subscription module explained the delivery schedule and the savings, and never said a shopper could skip, pause, or cancel. Risk-sensitive buyers read the silence as a trap. They are not wrong to.
The hidden cost of a defaulted subscription
Open the subscription module, default it, or both? The conversion answer and the trust answer point in opposite directions, and mobile is where the difference gets expensive.
The ecommerce bugs your funnel dashboard will never catch
A link labelled Reviews that opened the FAQ. An abandoned cart email whose product image was not clickable. Neither throws an error, neither appears in a report, and both cost money every day they survive.
Your best looking product page may still be losing buyers
Polish and clarity are different achievements. A page can win on art direction and still fail to tell a buyer whether the thing fits, how it arrives, or what happens if it is wrong.
Your reviews are not one pool. Treating them like one costs conversions.
Reviews for several different models were pooled into a single feed. The rating looked excellent and the content was useless to the buyer reading it, which is the worst of both worlds: it looks like proof and works like noise.
Simulate Before You Act: What Frontier AI Is Teaching DTC Brands About Conversion
World Labs just proved the thesis. Here's what it means for the brands spending real money on ads they haven't tested yet.
Similarweb 2026 Data: AI Search, The D2C Funnel, and Where It’s Reassembling
AI is transforming traditional search as buyers’ Step 1, and the funnel it created is invisible to attribution built for the old one.
Tagging reviews by the problem the buyer is trying to solve
A buyer with back pain and a buyer who sleeps hot are not looking for the same evidence, even on the same product. Letting reviewers say which problem they were solving turns a rating into decision support.
When your subscription architecture decides what you are allowed to test
A brand runs its recurring orders through a zero-price product variant that gets swapped at fulfillment. It works. It also quietly rules out several of the pricing experiments the merchandising team most wants to run.
When your marketplace pricing makes the subscription discount impossible
A brand selling across a marketplace, big-box retail, comparison shopping and its own site cannot simply cut the subscription price. The discount you can advertise is constrained by every other channel. Here is what to sell instead.
Your customer data is what makes a simulated buyer resemble a real one
Research on synthetic respondents is clear about where they hold up and where they break. The difference is almost always specification, and the best specification input a brand has is its own customer data.
You do not need more ideas. You need to know which one to test first.
Most teams have a backlog of thirty page improvements and capacity for four. Two thirds of tested ideas fail to move the metric they targeted. Choosing well matters more than testing more.
Cannes Lions 2026: The Year Agentic Commerce Enters the Chat
Live from the Croisette, the AI conversation has moved past creative tooling into agentic commerce — autonomous systems that decide and transact for consumers. Here's what it means for brands, and why understanding the buyer before the agent does is now the work.
Cannes Lions 2026: The Missing Conversation About AI and Creative Effectiveness
Every brand and agency at Cannes is asking how AI changes creative. Almost none are asking whether their creative actually converts the buyer it was made for. That’s the gap eLLMo closes.
Your buyer is becoming an agent — and four 2025 studies show what that does to conversion
The first real evidence is in. When people delegate decisions to AI agents, markets don't get cleaner — they inherit the human and the model behind every prompt. Here's what that means for anyone whose revenue depends on a yes on a page.
The gift-buyer conversion gap: why your DTC product page is speaking to the wrong customer
Most DTC product pages are written for self-purchasers. But for hundreds of brands, the gift-buyer is the highest-converting segment — and the most underserved. Here's what eLLMo Simulation consistently finds.
5 things A/B testing can't tell you before you spend
A/B testing is valuable. It's also backward-looking, slow, and expensive. Here are five things it can never tell you — and what pre-spend simulation gives you instead.
What OCEAN has to do with why your hero image isn't converting
The OCEAN (Big Five) personality model has been used in psychology for decades. eLLMo applies it to synthetic buyer personas — and it changes what we find in conversion analysis. Here's how each dimension shapes what a buyer sees, feels, and does on your landing page.
BFCM Conversion Prep: Why Smart DTC Brands Run Simulations 90 Days Before Black Friday
Black Friday is the highest-stakes conversion event of the year — and most DTC brands walk into it with no idea what's wrong with their page. Here's the pre-spend simulation playbook that top DTC operators are using to de-risk their BFCM spend.
The Trust Signal Hierarchy: What Skeptical Buyers Look for Before They Convert on a DTC Page
Not all trust signals are created equal — and some actively backfire with high-skepticism buyers. eLLMo Simulation consistently surfaces a trust signal hierarchy that most DTC brands get backwards. Here's what actually works.
See it on your page
Run eLLMo Simulation on your actual landing page, live, during a 30-minute demo.