The Age of AI Recommendations — Can Your Company’s Product Pages Pass AI Screening?

Recently, at a meeting in Tokyo, our SEO specialist partner Takeshi Gunji, CEO of CONTENTS SEO LAB, gave me a briefing on the latest developments in AI and SEO. The 90 minutes flew by, and what follows is my best effort to share what I've learned.

2025 was the year AI shifted decisively from "handy search tool" to "purchasing decision partner."

OpenAI added shopping capabilities to ChatGPT in April 2025. By September, they launched "Instant Checkout" — completing purchases without leaving the chat. Then in November, they introduced Shopping Research, a feature that interviews users about their budget, use case, and priorities, then cross-references multiple sources to generate a personalized buyer's guide.

AI no longer just returns search results. It researches products on the user's behalf, compares options, and recommends: "This one fits your situation. Here's why."

This report, based on research by our partner CONTENTS SEO LAB, examines how AI is reshaping purchasing behavior and what SMBs need to do about their information architecture — now.

Chaos — Data Reveals a Quiet Revolution

B2B Purchasing — AI Is Already the Trusted Advisor

Multiple surveys from the Japanese market paint a clear picture of AI's penetration into B2B purchasing.

An Ioix (SEO Japan) survey (August 2025, n=437) found that 90.6% of B2B product selection stakeholders use web search for information gathering — still the top channel — but AI search has already reached 43.5%. Around 40% of Google searchers also reference AI Overviews (Google's generative AI summaries).

LANY's survey (August 2025, n=110) goes deeper. 45.4% of respondents relied on AI for over 60% of their service evaluation process, and 46.4% said they discovered and ultimately contracted a vendor they hadn't previously considered — through AI. 87.3% said AI recommendations influenced their contract decisions, and satisfaction with AI-recommended services hit 92.8%.

The trust hierarchy is telling. In the same survey, vendor websites ranked first (26.4%), comparison sites second (22.7%), and generative AI third (20.0%) — above vendor sales reps (16.4%).

A TRENDEMON JAPAN survey (October 2025) confirmed that 91.9% of B2B marketers at companies with ¥50M+ revenue or 200+ employees use generative AI for competitive product research.

Then there's the NK Energy System survey (January 2026, n=180), which quantifies a tectonic shift in the sales relationship. 73.9% said internal evaluation proceeded without a sales rep present. 72.8% of final decision-makers chose their vendor before speaking with all candidates' sales teams. The top value buyers seek from AI: "Help me figure out if this fits my company" (50.0%), followed by "Organize comparison criteria" (38.9%).

B2C Purchasing — AI Shopping Has Gone Mainstream

The shift is equally dramatic in consumer markets.

BCG's holiday shopping survey (October 2025, 10 countries, n=10,240) found that 48% of consumers had already used or planned to use generative AI for holiday shopping — up 9 points year-over-year. Adoption has spread well beyond early adopters, reaching GenX (42%, +8pt YoY) and Baby Boomers (31%, +7pt YoY).

Adobe Digital Insights (July 2025) reported that AI-referred traffic to retail sites grew 4,700% year-over-year. AI-referred visitors showed 10% higher engagement and 32% longer session times.

In Japan specifically, BCG's Global Consumer Radar shows generative AI usage reaching 48%, tripling from 16% in September 2023.

Search Behavior Itself Has Changed

Traditional search queries were keyword fragments: "Omotesando BBQ." AI-era searches are conversational and specific: "Kid-friendly yakiniku restaurant in Omotesando with good kalbi."

This means product and service information can no longer just "match keywords." AI needs structured information to determine "this product fits this person's situation."

Reality — Why Most Product Pages Fail the AI Test

Research by CONTENTS SEO LAB revealed an uncomfortable truth: most companies' product pages fail the AI recommendation test.

Common Failure Patterns

Specs are listed, but "value to whom" is missing. Consider a materials manufacturer listing "water resistance" as a product feature. But "water resistance" means completely different things to different buyers. For a beverage label maker, it means "labels survive condensation — zero returns." For an election poster printer, it means "posters survive typhoon season." Same spec, entirely different value.

This reveals a core principle:

Context × Perspective = Value

The same spec delivers different value depending on the buyer's context (industry, challenge, use case) and perspective (what they're trying to achieve). Yet most product pages stop at "water resistant." AI can't determine who to recommend this product to with that information alone.

"Who it's NOT for" is never stated. Companies naturally want to sell to everyone. But the information AI trusts is trade-off disclosure — "this service isn't a good fit for these types of users." In LANY's survey, the top reason buyers trust AI was "it tells me the risks and caveats that sales reps hide" (45.5%). Honesty is the ultimate competitive advantage in the AI era.

FAQs cover logistics, not decisions. "How much is shipping?" "Can I return it?" That's not what AI needs. AI needs: "For a 30-person manufacturing company with a ¥50,000/month budget, does this tool deliver ROI?"

Reviews aren't connected to value propositions. Reviews exist, but they're not structured to show which value proposition each user segment is validating. AI can't use unstructured reviews as recommendation evidence.

Numbers — Information Architecture for Getting Chosen by AI

Step 1: Redesign Your Content's Information Architecture

The first priority is restructuring product and service pages. Here's the framework validated through research by CONTENTS SEO LAB.

1. One-line summary + 3 key differentiators (30–60 characters + 3 points)
This is what AI reads first. Replace vague copy with specific, differentiated claims.

2. Use case scenarios
Not abstract "use cases" — concrete situational descriptions. "For weeknight dinner prep." "For internal training at a 50-person IT company." "For an ecommerce site doing $700K annual revenue."

3. Persona-based value mapping
For each target user segment, explicitly state: the value they gain (benefits), the trade-offs they accept (limitations), and the fit criteria (checklist). This is the "Context × Perspective = Value" principle made tangible.

4. Decision-making FAQs
Not "How much does this cost per month?" but "For a 20-person marketing team, can this investment pay for itself in 3 months?"

5. Trust signals
Awards, third-party reviews, manufacturing transparency, specific implementation track records.

Step 2: Hand AI Your "Business Card" with Structured Data

Once the information architecture is right, the next step is making it machine-readable. Enter structured data.

Structured data is a standardized format for describing web page content so search engines and AI can understand it easily. The most common format is JSON-LD (JavaScript Object Notation for Linked Data).

Think of it as your product page's "business card" for machines. A human reading the page might understand "this is an SMB marketing tool at $200/month." AI might not. Structured data explicitly tells AI: "This is software, the target audience is companies with fewer than 200 employees, the price is $200/month, and these are its key attributes."

The technical implementation belongs to your development team, but here's what business owners and marketers need to know.

What structured data communicates
Beyond basics (name, description, price), you can describe target users, use cases, and unique attributes per persona. It's machine-readable documentation of "who gets what value."

Why it matters now
Google's AI Overviews and ChatGPT's Shopping Research both reference structured data when building recommendation shortlists. Without it, your product may never make it onto AI's candidate list.

The SMB opportunity
According to CONTENTS SEO LAB's analysis, even large enterprises have barely begun implementing structured data on product pages. SMBs that move early can leapfrog larger competitors in AI recommendations.

Step 3: E-E-A-T — Building Layers of Trust

If structured data is the entry ticket to AI's candidate list, EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) determines your ranking within that list.

Both Google and OpenAI are evolving toward trust-weighted recommendations. Even when AI recommends a product, users are more likely to click on brands they've heard of before.

Sustained media presence — owned content, social media, industry publication contributions — builds the brand recognition that maximizes the impact of AI recommendations.

Merits and Drawbacks of This Approach

✓ Why You Should Do This

SEO and AI optimization are two birds, one stone. Better information architecture improves traditional Google rankings AND AI recommendation probability. The investments align — no wasted effort.

SMBs can capture first-mover advantage. Even enterprises haven't addressed this yet. Moving early means you can secure the "AI-recommended" position before larger competitors catch up.

Conversion quality improves. AI-referred visitors already understand whether your product fits their needs. LANY's 92.8% satisfaction rate reflects AI pre-qualifying fit before the user even arrives.

Sales efficiency in the post-rep era. When 73.9% of internal evaluations proceed without a sales rep, your product page's information quality IS your sales team.

✗ The Hard Parts

Stating "who it's NOT for" takes courage. Traditional marketing defaults to "sell to everyone." Disclosing target limitations feels like leaving money on the table.

Implementation costs are real. Adding structured data and modifying your CMS requires technical resources. A $20/month tool alone won't cut it — you may need an external partner.

Timeline is unpredictable. AI crawler indexing schedules (CommonCrawl, Googlebot, etc.) are outside your control. "Publish and immediately appear in AI" isn't how it works.

Measurement is still immature. There's no standardized tool for accurately tracking "how often AI recommends me." You can track AI-referred traffic via Google Search Console and analytics, but recommendation volume and ranking visibility remain open challenges.

Conclusion — Specs Tell AI "What It Is." Value Tells AI "Who It's For."

The data points to one simple fact:

AI doesn't care about marketing copy. AI wants structured, honest information about who this product fits, what they'll gain, and what they'll have to give up.

Here's what SMBs can start doing today.

First, ask ChatGPT about your product. Type "your product name" — who should use this?" If AI can't give an accurate, specific answer, your page doesn't have enough information. I did give it a go, and it gave an answer which will be suitable for our previous version (we updated our site a week ago)

Next, articulate value per persona. The same product delivers different value to User A and User B. Making that difference explicit improves AI's recommendation accuracy.

Then, make honesty your competitive edge. Writing "This service isn't a great fit for X type of company" feels like lost revenue in the short term. But in the AI era, trade-off disclosure builds trust, and trust drives recommendations.

This shift is accelerating. Japan's generative AI usage tripled from 16% (2023) to 48% (2025). 73.9% of B2B evaluations proceed without sales reps. 46.4% of buyers contracted a service they discovered through AI.

Waiting until next month to start may already be too late. And yes — we're guilty of this gap on our own product pages too 😅.

One final note: this topic is too large for a single article. In an upcoming piece, we'll explore how AI is reshaping the role of "sales" itself in this new era.

✓ Try This If You...

• Sell B2B or B2C products/services online
• Have product pages that are spec-heavy but lack persona-based value propositions
• Want more visibility in AI shopping assistants and AI Overviews
• Have limited sales headcount and need content to carry the sales load

✗ Skip This If You...

• Haven't yet established product-market fit
• Sell entirely offline/in-person with no digital touchpoints
• Don't have a website yet (build that first)