Skip to main content
Models: 9
Dimensions: 26
Trials: 56,640
Pre-registered: osf.io/et4nf

Web Benchmark Analysis

A comprehensive analysis of 213 consumer product pages across 7 categories, measuring their Machine Likeability scores.

Executive Summary

213
Pages Analyzed
49.6
Average Score
40
Minimum Score
81.3
Maximum Score

This benchmark represents the most comprehensive analysis of Machine Likeability across real-world product pages to date. We analyzed 213 pages from leading brands across 7 major product categories, measuring their optimization across all 26 AI preference dimensions.

The results reveal a massive optimization gap in the market. The average web page scores just 49.6 out of 100 for Machine Likeability, with scores ranging from 40 to 81.3. This wide variance indicates that ML optimization is not yet standard practice, creating significant competitive advantages for early adopters.

Key Insight

The average web page scores just 49.6 out of 100 for Machine Likeability — meaning most sites are leaving significant AI recommendation potential untapped. Even leading brands from Google (40.5), Linear (40.0), and Paula's Choice (40.0) fail to implement basic optimization signals.

Category Performance Analysis

Telecom (55.4 avg, Top Performer)

Leader

Telecom companies lead with an average score of 55.4, driven by strong bundle offerings and clear pricing structures. T-Mobile's home internet page (81.3) sets the benchmark with exceptional third-party authority signals, detailed plan comparisons, and regulatory transparency.

Electronics (51.0)

Electronics retailers average 51.0, with gaming brands like Razer (72.1) excelling through detailed product specifications and bundle options. The category succeeds with specification-heavy content.

Software (49.4)

Software companies average 49.4. Modern SaaS design prioritizes minimalism over comprehensive information display, leaving AI systems with insufficient data for recommendations.

Apparel (44.5, Bottom)

Apparel is the lowest-performing category. DTC brands struggle to communicate value in AI-readable formats. Human-trust signals don't translate to machine-readable confidence indicators.

Signal Presence Analysis

Measuring how often each of the 26 AI preference dimensions appears across the 213 analyzed pages.

Most Present Signals

Novelty Seeking (77%)77%

Most pages communicate what's new about their products. This is table stakes — brands understand the importance of positioning products as current and innovative.

Specificity Preference (69%)69%

Detailed specifications are common, especially in electronics and technical categories. Products with measurable attributes naturally include spec sheets.

Information Depth (66%)66%

Pages generally provide adequate detail. Most brands recognize that customers need information to make decisions, though depth varies by category.

Recommendation Revision (61%)61%

Many pages have content that could trigger recommendation changes — highlighting unique benefits or addressing specific use cases that differentiate products.

Bundle Preference (53%)53%

About half of pages mention bundles or packages. More common in telecom and electronics; rare in apparel and personal care.

Most Missing Signals

Ethical Concern (3%)3%

ESG and ethical sourcing signals are almost entirely absent. Even brands with strong ethical practices fail to communicate them in AI-readable formats.

Local Preference (6%)6%

Almost no pages mention local sourcing or production. This represents a massive missed opportunity for brands with local manufacturing stories.

Negative Review Weight (14%)14%

Very few pages address potential concerns proactively. Brands avoid mentioning limitations, missing the opportunity to build trust through transparency.

Social Proof (17%)17%

Despite being the #1 AI selection driver in research, only 17% of pages have visible social proof. Reviews exist but aren't prominently displayed or are hidden behind clicks.

Sustainability (17%)17%

Environmental messaging remains rare despite growing importance to both consumers and AI recommendation systems. Even eco-focused brands often fail to quantify impact.

The Signal Gap

54%

The average page is missing 54% of the signals AI models look for.

This represents a massive optimization opportunity. Early adopters can gain significant competitive advantages by addressing these gaps.

We measured the "target signal strength" for each dimension based on top-performing pages, then calculated how many pages reach that target. The results are sobering:

DimensionPages at TargetPercentAverage Gap
Bundle Preference24 / 21311%47 points
Specificity Preference18 / 2138%51 points
Third-Party Authority12 / 2136%67 points
Sustainability8 / 2134%83 points
Social Proof0 / 2130%83 points
Local Preference0 / 2130%94 points
Ethical Concern0 / 2130%97 points

What This Means

Only 24 of 213 pages (11%) hit the target for Bundle Preference — the best-performing dimension. For Social Proof, ZERO pages hit the target, despite social proof being the #1 driver of AI recommendations in our research. Even pages with reviews don't display them prominently enough for maximum AI impact.

For Ethical Concern and Local Preference, virtually no pages even attempt these signals. This creates a blue ocean opportunity: brands that authentically communicate ethics and locality can dominate AI recommendations in those dimensions with minimal competition.

Model Consensus Analysis

All six AI models show remarkably similar scoring patterns, with averages ranging from 44.6 (Perplexity) to 49.1 (Gemini). This consensus suggests the signals measured are truly universal across AI systems, not model-specific quirks.

Cross-Model Consistency

The 4.5-point spread between the most lenient (Gemini, 49.1) and strictest (Perplexity, 44.6) models is remarkably small given the diversity of architectures and training approaches. This consistency validates our research: these signals represent fundamental patterns in how AI systems evaluate product information, not artifacts of specific model implementations.

GPT-5.4, Claude, and Gemini cluster tightly around 48-49, suggesting similar training on commercial content evaluation. O3's reasoning capabilities don't significantly change its scoring (47.2), indicating that these signals work at the system level, not just for quick-response models.

Perplexity: The Strict Evaluator

Perplexity is the strictest evaluator at 44.6 average, likely due to its search-oriented training emphasizing factual grounding and source attribution. Pages that score well with Perplexity tend to have exceptional third-party authority signals and detailed specificity.

The Perplexity-Gemini spread of 4.5 points means your score should be relatively consistent regardless of which AI recommends you. If you optimize for the strictest model (Perplexity), you'll perform well across all platforms. Conversely, a page scoring 40 with Gemini will still score poorly (~36) with Perplexity — there's no gaming the system.

Strategic Implication

Because all models show similar patterns, you don't need separate optimization strategies for different AI platforms. Focus on the fundamental signals — social proof, specificity, third-party authority, sustainability — and your improvements will translate across all AI recommendation systems. This makes ML optimization more approachable: one comprehensive improvement benefits all channels.

Top 10 Performers: What Makes Them Succeed

Detailed analysis of the highest-scoring pages, examining specific signals and extracting lessons for other brands.

#1

t-mobile.com

Telecom

https://www.t-mobile.com/home-internet

81.3
ML Score
View Report →

Why it succeeds:

  • Strong bundle preference signals with clear package comparisons
  • Third-party authority with visible awards and certifications
  • Clear pricing anchors and transparency
  • Excellent specificity in service details

Key Lesson:

Telecom's regulatory requirements for transparency actually help AI readability. The combination of bundle options, third-party validation, and specific service details creates a gold standard for ML likeability.

#2

razer.com

Electronics

https://www.razer.com/gaming-laptops

72.1
ML Score
View Report →

Why it succeeds:

  • Exceptional product specification detail
  • Clear bundle and configuration options
  • Strong comparison framing across models
  • Technical authority signals throughout

Key Lesson:

Gaming brands excel because they naturally speak in specifications. The detailed technical data that gamers demand is exactly what AI models need to make confident recommendations.

#3

soylent.com

Food & Beverage

https://soylent.com/products/soylent-drink

70.2
ML Score
View Report →

Why it succeeds:

  • Strong sustainability messaging
  • Detailed nutritional specificity
  • Clear use case and novelty positioning
  • Social proof through community engagement

Key Lesson:

Soylent succeeds by addressing multiple dimensions: sustainability appeals to value-based signals, nutrition specs hit information depth, and community presence provides social proof.

#4

ikea.com

Home Goods

https://www.ikea.com/us/en/p/poang-armchair

68.5
ML Score
View Report →

Why it succeeds:

  • Excellent product specificity (dimensions, materials)
  • Strong sustainability signals
  • Clear assembly and warranty information
  • Price anchoring with family product comparisons

Key Lesson:

IKEA's focus on practical details (measurements, materials, assembly) combined with sustainability messaging creates a comprehensive AI-optimized experience.

#5

github.com

Software

https://github.com/features/copilot

68.8
ML Score
View Report →

Why it succeeds:

  • Strong recommendation revision potential
  • Clear use case specificity for developers
  • Third-party authority through GitHub brand
  • Detailed feature comparisons

Key Lesson:

GitHub Copilot leverages its platform authority and developer-focused specificity. Technical products benefit from detailed feature explanations.

#6

verizon.com

Telecom

https://www.verizon.com/5g/home-internet

67.9
ML Score
View Report →

Why it succeeds:

  • Clear bundle and plan comparisons
  • Strong pricing transparency
  • Third-party authority signals
  • Specific coverage and speed details

Key Lesson:

Another telecom winner. The pattern is clear: regulatory transparency + bundle options + specific technical details = high ML scores.

#7

wayfair.com

Home Goods

https://www.wayfair.com/furniture/pdp/wade-logan-sectional

66.3
ML Score
View Report →

Why it succeeds:

  • Extensive product specifications
  • Strong social proof (reviews, ratings)
  • Clear comparison with similar items
  • Detailed return and warranty information

Key Lesson:

Wayfair shows how massive review volumes and detailed specs can overcome commodity product challenges. Social proof at scale works.

#8

apple.com

Electronics

https://www.apple.com/macbook-pro

65.7
ML Score
View Report →

Why it succeeds:

  • Exceptional product specificity
  • Strong brand authority
  • Clear configuration options
  • Environmental sustainability messaging

Key Lesson:

Apple's ML score comes from technical precision and environmental messaging, not social proof. Brand authority can partially compensate for missing review signals.

#9

att.com

Telecom

https://www.att.com/internet/fiber

64.8
ML Score
View Report →

Why it succeeds:

  • Clear plan and bundle comparisons
  • Pricing transparency
  • Specific speed and service details
  • Installation and setup information

Key Lesson:

The third telecom in the top 10 confirms the pattern. Category leaders emerge when industry norms align with AI preferences.

#10

dell.com

Electronics

https://www.dell.com/en-us/shop/dell-laptops/xps-15

63.2
ML Score
View Report →

Why it succeeds:

  • Detailed technical specifications
  • Clear configuration and customization
  • Comparison across models
  • Business-focused authority signals

Key Lesson:

Dell succeeds through comprehensive technical detail and business authority positioning. B2B signals can be as powerful as B2C social proof.

Bottom 10 Performers: Understanding The Gap

Common Pattern

The bottom 10 share a consistent pattern: zero social proof, zero third-party authority, and minimal specificity. Critically, these aren't bad products — Blue Bottle Coffee, Linear, and Paula's Choice are all market leaders with loyal customers and strong brands. They simply haven't optimized for AI visibility. This demonstrates that brand strength and product quality don't automatically translate to ML scores.

#213

bluebottlecoffee.com

Food & Beverage

https://bluebottlecoffee.com/coffee

40
ML Score
View Report →

Missing signals:

  • Zero social proof (no reviews or ratings visible)
  • No third-party authority signals
  • Minimal product specificity beyond origin
  • No sustainability messaging despite premium positioning
  • Missing comparison framing

Pattern:

Premium brand relying entirely on aesthetic presentation and brand name, with no AI-readable signals.

#212

linear.app

Software

https://www.linear.app/pricing

40
ML Score
View Report →

Missing signals:

  • No social proof or customer testimonials
  • Missing third-party authority
  • Limited feature comparison detail
  • No use case specificity
  • Minimal information depth

Pattern:

Modern SaaS design prioritizing minimalism over AI discoverability. Clean UI, invisible to AI.

#211

paulaschoice.com

Personal Care

https://www.paulaschoice.com/skin-perfecting-bha-liquid

40
ML Score
View Report →

Missing signals:

  • Reviews exist but not prominently displayed
  • No third-party authority (despite research backing)
  • Missing sustainability or ethical signals
  • Limited comparison framing
  • Weak specificity in ingredient benefits

Pattern:

Science-backed brand failing to communicate research authority. Data exists but not AI-accessible.

#210

allbirds.com

Apparel

https://www.allbirds.com/products/mens-wool-runners

42
ML Score
View Report →

Missing signals:

  • Weak social proof presentation
  • Sustainability mentioned but not quantified
  • Minimal product specificity
  • No comparison framing
  • Missing material detail depth

Pattern:

DTC darling with strong sustainability story not optimized for AI parsing. Marketing speaks to humans, not machines.

#209

glossier.com

Personal Care

https://www.glossier.com/products/boy-brow

42.3
ML Score
View Report →

Missing signals:

  • Minimal product specifications
  • No third-party validation
  • Missing ingredient transparency
  • Weak comparison options
  • Limited use case detail

Pattern:

Instagram-native brand optimized for visual appeal, not AI comprehension. Strong community, weak signals.

#208

everlane.com

Apparel

https://www.everlane.com/products/mens-organic-cotton-tee

43.1
ML Score
View Report →

Missing signals:

  • Pricing transparency present but other signals missing
  • Weak social proof display
  • Limited material specificity despite "transparent" positioning
  • No comparison framing
  • Sustainability claims not quantified

Pattern:

Radical transparency in pricing not extended to product details. Human-trust signals not machine-readable.

#207

warbyparker.com

Apparel

https://www.warbyparker.com/eyeglasses/men/percey

43.8
ML Score
View Report →

Missing signals:

  • Try-at-home program not recognized as risk mitigation
  • Weak social proof presentation
  • Limited product specificity
  • No material or construction detail
  • Missing comparison framing

Pattern:

Innovative customer experience (home try-on) not translating to AI-readable signals.

#206

allmodern.com

Home Goods

https://www.allmodern.com/furniture/pdp/sectional

40.2
ML Score
View Report →

Missing signals:

  • Generic product descriptions
  • Weak social proof
  • Minimal material specificity
  • No sustainability information
  • Limited comparison options

Pattern:

Budget furniture site with commodity products lacking differentiating details.

#205

store.google.com

Electronics

https://www.googlestore.com/product/pixel_8

40.5
ML Score
View Report →

Missing signals:

  • Surprisingly weak social proof for major brand
  • Limited third-party validation
  • Weak comparison framing
  • Minimal environmental detail
  • Missing bundle optimization

Pattern:

Even Google fails basic ML optimization. Brand confidence leading to signal complacency.

#204

huel.com

Food & Beverage

https://www.huel.com/products/huel-black-edition

44.2
ML Score
View Report →

Missing signals:

  • Nutritional detail present but not contextualized
  • Weak social proof presentation
  • Limited sustainability quantification
  • Missing comparison framing
  • No third-party validation

Pattern:

Data-rich product not presenting data in AI-optimized format. Numbers without narrative.

The Good News

None of these issues are structural — they're all fixable with content updates. Blue Bottle could add customer testimonials and origin story details. Linear could showcase customer logos and use cases. Paula's Choice could display clinical research and dermatologist endorsements. The brands already have this information; they just need to make it AI-accessible. A page scoring 40 today could reach 65+ with strategic content additions that don't require product changes or major redesigns.

Actionable Insights: What To Do Next

1

Add Social Proof

83% of pages lack visible reviews/ratings. If you have reviews, display them prominently above the fold. Include star ratings, review counts, and specific testimonials. If you don't have reviews yet, start with customer logos, case study quotes, or social media mentions. Even basic social proof beats none.

Quick win: Add "4.8 stars from 1,247 customers" to your hero section if you have it.
2

Show Third-Party Authority

Only 33% mention awards, certifications, or press. If you've won awards, earned certifications, or been featured in media, display those signals. AI models weight external validation heavily. Industry certifications, "As seen in..." press logos, and award badges all count.

Quick win: Add a trust badge section with any press mentions, certifications, or awards.
3

Communicate Bundle Options

47% miss bundle/package signals. If your product can be combined with others, make that clear. "Frequently bought together," "Complete the set," or "Bundle & save" sections all trigger bundle preference signals. Even suggesting complementary products counts.

Quick win: Add "Buy with [complementary product] and save 15%" to product pages.
4

Address Sustainability

83% have no environmental messaging. If you have any sustainability practices — recycled materials, carbon offset shipping, energy-efficient production — mention them. Quantify when possible: "Made from 80% recycled materials" beats "eco-friendly."

Quick win: Add a sustainability section even if it's basic: "Carbon-neutral shipping on all orders."
5

Include Local Signals

94% miss local production/sourcing. If you manufacture locally, source locally, or have local roots, say so explicitly. "Made in Portland, Oregon" or "Locally sourced ingredients" are powerful differentiators with almost zero competition in this dimension.

Quick win: Add location information if applicable: "Handcrafted in Brooklyn since 2018."

Ready to Optimize Your Machine Likeability?

Use our ML Score tool to analyze your product pages and get specific, actionable recommendations for improving AI discoverability.

Score Your Page Now