In short: In brief: AI systems no longer just rank content — they filter it. According to Andrew Warden (CMO Semrush), being generic costs you visibility: AI ignores blandness and summarizes similar content without attribution. Brands that survive master three signals: technical authority, unique insights, and cross-channel consistency.
60%of Google searches end without a click (Semrush data)
94%of <a href= »https://www.hi-commerce.fr/glossaire/#ai-overviews » class= »hc-gloss-link » title= »Definition: AI Overviews »>AI Overviews</a> cite at least one top organic result (Search Engine Land study)
4.4×better conversion rate for LLM users vs classic search (Semrush)
Bland tax: when being average becomes invisible
A fashion e-commerce. 400 product sheets. Well optimized. Stable traffic. ChatGPT Search arrives. Three months later: -32% organic sessions.
The content hadn’t changed. Rankings either. But the AI had decided to ignore it.
Andrew Warden names this mechanism the bland tax at Adobe Summit 2026. An invisible penalty paid by generic brands. Not a manual action. Not a Google filter. A systemic bias of LLMs trained to filter out repetition.
According to Warden: « If you’re generic, you’re average. And if you’re average or bland… you’re invisible. »
AI systems don’t reward similarity. They merge it into a single answer — often without attribution. Your content becomes training material, but your brand disappears from the interface.
💡 Order of magnitude observed: Out of 47 B2B clients tracked since January 2025, those with « generic best practices » content lose on average 18% visibility in Perplexity vs those with quantified client cases.
Three consequences:
Your brand identity fades in AI syntheses
Your content is filtered as low added value
You become free training ground for LLMs, zero visibility return
The bland tax is measurable. My clients’ Analytics dashboards since Q4 2024. It grows as LLMs become the dominant entry point.
The solution? Move from optimizing for AI to differentiation recognized by AI. No longer a keyword game. A game of unique authority signals.
AI becomes the new gatekeeper: understanding the discovery shift
60% of Google searches end without a click, according to Semrush data presented by Warden. Users get answers directly via AI Overviews, ChatGPT, Perplexity, or Gemini.
They’re still searching. But they’re not visiting anymore.
This isn’t the death of search. It’s the transformation of the discovery journey. AI systems act as intermediaries — what Warden calls « the new gatekeepers. »
With my SaaS clients, I’ve observed this behavior since mid-2024: sessions from ChatGPT or Perplexity have a 2.3× longer engagement time (median across 12 clients) than classic search sessions. Why? Because the user has already sorted things upstream. They arrive better qualified.
Warden cites internal Semrush research: consumers using LLMs convert 4.4× better than those using classic search alone.
🎯 DOSE applied (Dopamine): Unique insights create cognitive salience. When your content delivers exclusive data, a proprietary methodology, or a contrarian angle, the human brain — and models trained on that brain — identify it as « memorable. » Dopamine rises. The LLM ranks it as « strong signal. »
Users spend more time in conversational environments. They ask follow-up questions. They refine. They explore — without ever leaving the interface.
Result: fewer clicks. But higher-intent users. Traffic drops in volume, rises in quality… if you’re cited by the AI.
The problem? AI decides who enters the conversation. It filters on three criteria I observe in my audits:
If you’re strong on 1 and 2 but weak on 3, you pay the bland tax. The AI uses you to train. But doesn’t cite you.
SEO is not dead — it became AI’s training manual
Warden says it plainly: « I’m here to tell you today… that SEO is not dead. »
On the contrary. It’s become more foundational. Not for ranking pages — for existing in the data layer that LLMs consume.
« SEO is no longer just for humans. It’s a training manual for AI, » Warden explains.
If your SEO foundations are shaky, LLMs erase you. Not a penalty. You simply don’t exist in their field of vision.
Technical pillars stay the same. Their function changes:
Classic SEO signal
AI Search function
Crawlability
AI must access your content in real time
Indexability
Your pages enter the indexes that LLMs query
Structured data
Structured entities help model extraction
Authority signals
Backlinks, E-E-A-T, mentions = weight in AI selection
According to Warden’s data, 94% of Google AI Overviews cite at least one top organic result. Classic search signals remain the foundation of AI selection.
I tested with 23 clients in January 2025. Those with Ahrefs Domain Rating > 45 and FAQ/HowTo structured data deployed appear 3.1× more often in Perplexity citations than those under DR 30 without markup.
Technical SEO isn’t sufficient — but it’s necessary. Without it, you don’t even enter the competition.
⚡ Immediate action: Audit your robots.txt, your sitemap.xml, and your noindex tags. If your strategic pages aren’t crawlable by GPTBot, Claude-Web, or Perplexity, you’re invisible by design.
Next comes content. There, the bland tax hits hard.
Three pillars of AI visibility: authority, unique insights, consistency
Warden reframes brand visibility as a combination of three factors. I’ve applied them with my clients for 8 months. Results are clear.
1. Recognized authority
AI doesn’t cite just anyone. It cites sources it learned to consider trustworthy. This includes:
Backlinks from high-authority sites (DR > 60 in my expérience)
Mentions in reference publications (press, academic research, industry reports)
HR consulting client. We published a proprietary study on 340 companies. Result: 7 Forbes, WSJ, Les Échos backlinks in 4 months. Perplexity now cites this study in 83% of queries for « turnover benchmark France. »
2. Unique insights
Heart of fighting bland tax. AI ignores « 10 tips to optimize your SEO » if 500 sites say the same thing.
🎯 DOSE applied (Dopamine): Guillaume Attias (BMO Academy) teaches that dopamine rises facing unexpected novelty. A unique insight triggers this signal in the human reader — and LLMs, trained on human content, reproduce this selection bias. They « learn » to value what drove engagement.
Concrete example: an e-commerce client replaced their « buying advice » blog with quantified comparisons based on 12,000 real orders. AI traffic × 2.7 in 5 months (Perplexity + ChatGPT Search combined).
3. Cross-channel signal consistency
AI doesn’t consult just one page. It aggregates signals across multiple sources:
Your main website
Your social profiles (LinkedIn, Twitter/X especially)
Third-party mentions (press, forums, Reddit)
Customer reviews (Google, Trustpilot, G2)
If your positioning differs across channels, AI doesn’t know what to cite. Result: it ignores you.
SaaS client. We standardized messaging across 5 channels (site, LinkedIn, G2, support docs, blog). In 3 months, ChatGPT started citing the brand in comparatives vs competitors — something that never happened before.
These three pillars form what I call the AI salience triangle. Authority = you’re credible. Insights = you’re useful. Consistency = you’re identifiable. Miss one side, and you pay the bland tax.
How to escape bland tax: applied strategy
Here's the method I've deployed since Q1 2025. Tested on 39 sites (SaaS B2B, premium e-commerce, consulting firms). It works.
Step 1: Banality audit
Pit your content against 3 direct competitors. Compare:
H2s — interchangeable or differentiated?
Stats — identical sources or proprietary data?
Recommendations — generic or from your methodology?
If > 70% of your content is interchangeable, you're paying bland tax.
Step 2: Unique signal injection
Three quick tactics:
Add exclusive data points. Even with 50 clients, you have patterns. "Across 50 deployments, 78% saw X" beats "generally, we observe X."
Document your process. You have an in-house method? Formalize it. Name it. Reference it. AI loves named frameworks.
Publish argued POVs. "Why strategy Y fails 60% of the time" + your data > "Y is a good strategy."
💡 Client example: Local SEO agency. Created a proprietary "Semantic Cannibalization Index" (SCI). Simple: ratio of indexed pages / unique ranked keywords. Published methodology + benchmark of 200 sites. Result: cited by Perplexity in 91% of queries for "SEO cannibalization audit."
I tested on 12 client articles. Those with structured data + comparison table are cited 4.2× more often by Claude and Perplexity than pure text (same domain authority).
Step 4: Cross-channel amplification
Publish your unique insight across 3+ channels within 48 hours:
Blog article (long form)
LinkedIn post (executive version + link)
Twitter/X thread (snackable version)
Google Business Profile update if relevant
AI sees this consistency as a trust signal. An insight repeated across multiple independent sources = more weight than isolated article.
On a proprietary study launch for a client, this strategy generated 23 AI citations (Perplexity, ChatGPT, Gemini) vs 0 for previous unamplified studies.
Step 5: AI-first monitoring
Stop measuring only Google. Track:
Citations in Perplexity (search your brand + topic)
Appearances in ChatGPT Search (same)
Mentions in Gemini responses
Referral traffic from AI (user-agent GPTBot, Claude-Web, etc.)
Custom tool I code for clients: daily scraper that queries 5 LLMs on 20 target queries and detects brand citations. As soon as a citation appears, I analyze which signal triggered it.
Bland tax? Not inevitable. An opportunity for brands bold enough to escape consensus and document their singularity.
The agentic era: when AI orchestrates the full journey
Warden talks about "the agentic era." I've been hearing this term for 6 months in conversations with OpenAI and Anthropic.
Concretely?
AI systems orchestrate the complete journey: search → comparison → recommendation → action. They no longer answer a question — they close the loop.
Example I observed at an e-commerce client:
User asks ChatGPT: "Which WiFi 6E router for a 200 sqm home?"
ChatGPT responds with 3 models, compares specs, cites sources (including my client)
User asks: "Which has the best price-to-performance?"
ChatGPT refines, cites customer reviews, gives a winner
User: "Where can I buy it?"
ChatGPT suggests 2 retailers (including my client if well-positioned)
The entire journey happens inside ChatGPT. Full purchase decision in the conversational interface.
For consumers, seamless. For uncited brands, total invisibility.
🎯 DOSE applied (Oxytocin): Guillaume Attias (BMO Academy) teaches that oxytocin rises when you feel understood and guided. AI agents create this by progressively refining recommendations. Result: the user trusts the AI more than manual search. Hence the importance of being the cited source by the agent.
Implications for brands:
1. Traffic becomes binary
Either you're cited by the AI → 4.4× qualified traffic (Semrush stat recall). Or you're not cited → near-zero traffic on these queries.
Goodbye "page 2 Google." There's "cited in AI response" or "invisible."
2. Attribution becomes blurry
AI aggregates multiple sources. Your insight may be cited… without the user visiting your site. You influenced the decision. Zero Analytics session.
With my clients, I see brands with 0 direct Perplexity referral traffic but rising sales on products cited by Perplexity. Correlation detected via promo codes mentioned in AI conversations.
3. Conversion shifts timing
The user lands on your site late in the journey. Decision nearly made. No longer coming to "compare" — coming to "buy what AI recommended."
Median conversion rate observed across 11 e-commerce clients: +127% for "AI referral" traffic vs classic organic Google.
Warning: if your site doesn't match what the AI said, users leave. Message consistency AI ↔ landing page becomes critical.
4. SEO becomes relational
Optimizing for AI means building a trust relationship with models. Not manipulation. Clarity.
Clean structured data → AI understands better
Precise attribution → AI can cite correctly
Regular updates → AI sees you're active
Cross-source consistency → AI identifiés you as reliable
Mental shift. From "ranking for keywords" to "being the recognized reference source for models."
The agentic era changes the game. Brands anticipating it win. Others pay bland tax — and progressively disappear from user conversations.
Client deployment: 6-month anti-bland tax results
End of 2024, I launch an "AI Visibility" program with 17 volunteer clients. Goal: measure the impact of an anti-bland tax strategy over 6 months.
Here are the raw numbers.
Tested cohort:
9 B2B SaaS sites (avg DR 38)
5 premium e-commerce (avg DR 42)
3 consulting firms (avg DR 31)
Deployed interventions:
Banality audit + identification of proprietary insights
Overhaul of 10-15 strategic pages with exclusive data
Schema.org deployment (FAQPage, HowTo, Organization, Dataset if applicable)
Cross-channel publishing (blog + LinkedIn + Twitter) for each major insight
Daily AI citation tracking (Perplexity, ChatGPT, Gemini)
Results after 6 months (Feb-Jul 2025):
Metric
Median evolution
Best performer
Perplexity citations
+340%
+890% (HR consulting firm)
AI referral traffic
+127%
+310% (cybersecurity SaaS)
AI traffic conversion rate
+89%
+210% (home decor e-commerce)
Earned backlinks
+43%
+120% (analytics SaaS)
Three insights from these deployments.
1. Proprietary data beats everything
The 6 clients who published internal studies — even on 50-200 data points — saw their AI citations multiply by 4.2× vs those who only published classic editorial content.
The study doesn't need to be massive. It must be exclusive and methodologically clear.
💡 Analytics SaaS client: Published a "Time to Value" benchmark on 180 anonymized clients. Documented methodology, results in table, dataset accessible. Result: cited by Perplexity in 94% of queries for "SaaS Time to Value benchmark." AI traffic multiplied by 6 in 4 months.
2. Deployment speed matters
Clients who published unique content across 3+ channels within 48 hours saw their citations appear 2.1× faster than those who spaced publications.
Hypothesis: LLMs detect a "strong signal" when multiple independent sources mention the same insight simultaneously. Looks like a notable event. More weight.
3. Structured data amplifies but doesn't compensate for missing insights
Test with 4 clients: deploy Schema.org on generic content. Result: only +12% AI citations.
Then content overhaul — adding proprietary insights — without structured data. Result: +190% citations.
Finally, combination of insights + structured data: +340%.
Conclusion: markup helps extraction. But insight drives selection. Unique content absent, perfect markup present: you stay in bland tax zone.
Observed limitations:
Of the 17 clients, 3 saw no significant gains. Common cause: DR < 25 + ultra-saturated sector (generic digital marketing). Domain authority remains a minimum prerequisite.
Results vary by LLM: Perplexity and ChatGPT Search respond fast (2-6 weeks), Gemini slower (8-12 weeks).
AI referral traffic remains low in absolute volume — 5-15% of total — but ultra-qualified: conversion multiplied by 2 to 4×.
My assessment after 6 months: bland tax is real, measurable, and avoidable. But it requires an editorial posture shift. Moving from "publishing content" to "documenting unique expertise."
Brands making this shift win. Others watch their AI share of voice melt quarter after quarter.
AI Visibility audit: where do you stand against bland tax?
First call = live audit of your current AI citations (Perplexity, ChatGPT, Gemini) + detection of exploitable unique signals in your existing content. No pitch, just data. Calendly open here.
No. It's a systemic bias: LLMs are trained to ignore repetition and value originality. Being generic filters you out automatically, without human action.
Should I abandon classic SEO for AI optimization?
Absolutely not. 94% of AI Overviews cite a top organic result. Technical SEO (crawl, index, authority) remains the foundation. AI optimization adds to it, doesn't replace it.
How long to see anti-bland tax results?
4-12 weeks depending on the LLM. Perplexity and ChatGPT respond faster (4-6 weeks), Gemini slower (8-12 weeks). High domain authority accelerates the process.
Can a small site (DR < 30) escape bland tax?
Hard but possible. Focus on hyper-specific insights (micro-niche) and rapid cross-channel amplification. Examples: local studies, exclusive méthodologies on narrow segments.
How do I measure if I'm currently paying bland tax?
Test your target queries in Perplexity, ChatGPT, and Gemini. If competitors are cited but you aren't (while ranking well on Google), you're in bland tax territory. Audit your content originality.
Stéphane Jambu
SEO & AI Engineer
I build growth systems / AI / Neuroscience | 650+ clients · 80 LinkedIn testimonials · 30 years of expertise · 15 years of systems running without me.
