In short: In short: SEO success is no longer measured by organic session volume. Search Engine Journal documented a fundamental shift: AI engines cite or ignore, but no longer systematically drive clicks. I’ve observed this phenomenon across 47 clients since November 2025 — traffic drops, but incoming requests explode.
47 sitesanalyzed since November 2025 — traffic down, citations up
-38%average traffic observed on well-ranked B2B sites (Nov. 2025 – Mar. 2026)
+127%qualified incoming requests same period (order of magnitude)
The metric that no longer holds up
A client calls me on a Tuesday evening in February. B2B SaaS site, well-ranked for 18 months. 12,000 organic sessions per month in October 2025. 7,400 in February 2026.
-38% in four months.
He panics. Asked if Google penalized him, if his hosting failed, if someone modified the canonical tags. I checked: nothing broken. Positions stable. Brand queries up. Core Web Vitals green.
Then he says: « But we’re getting twice as many inquiries as before. »
There’s the problem: traffic no longer measures what it used to. AI engines (Perplexity, ChatGPT Search, Google AI Overviews, Bing Copilot) extract, rewrite, cite — but no longer systematically send the user to click your page.
You’re cited. Not visited.
Search Engine Journal documented this in an April 2026 article by Dan Taylor: « Content quality still matters, but it is no longer the deciding factor. » Quality is no longer enough. What matters is retrievability — your ability to be extracted, reused, cited in synthesized answers.
I see this shift across 47 sites since November 2025. Traffic declining. Citations rising. Qualified requests exploding.
The KPI has changed. We’re no longer playing volume. We’re playing authority.
What the numbers tell us
I’m not talking about projections. I’m talking about what I observe.
Across 47 sites I’ve tracked since November 2025 (B2B, technical e-commerce, specialized publishers), here’s what Google Analytics and Google Search Console show:
Average organic traffic: -38% between November 2025 and March 2026
Search Console impressions: +12% same period
Average positions: stable or slightly up (+0.4 average position observed)
Qualified incoming requests (forms, calls, demo requests): +127% (order of magnitude, all sectors combined)
Translation: Google still shows your pages, but users don’t need to click anymore. They ask ChatGPT or Perplexity. The AI engine synthesizes. Your content is read, extracted, cited — but not visited.
Result: you lose traffic, but gain perceived authority. The user sees your name in an AI response. They find you credible. They search for you directly.
Dan Taylor puts it this way in the SEJ article: « We’ve moved from authorship to retrieval. » We no longer measure how many people read your page. We measure how many AI systems reuse it.
Why the DOSE framework makes complete sense now
The DOSE framework (Dopamine, Oxytocin, Serotonin, Endorphin), taught by Guillaume Attias at BMO Academy, rests on a simple idea: high-performing content triggers specific neurochemical reactions.
In AI Search, oxytocin becomes the central molecule.
Oxytocin is recognition. Trust. The feeling of being cited, referenced, considered a reliable source. Before, you got this signal through backlinks, social shares, press mentions. Today, you get it when an AI engine cites your content as a source in a synthesized answer.
Concrete example: a textile e-commerce client (1,200-item catalog) restructured its blog in January 2026. We stopped producing generic listicles. We built 47 ultra-targeted technical pages: fiber composition, OEKO-TEX standards, international size correspondence charts, care guides by material.
Traffic dropped. But Perplexity now cites this client systematically on textile technical queries. Result: qualified prospects arriving saying « I saw you master OEKO-TEX standards. »
They never visited the site. They read an AI response that cited it.
That’s oxytocin in AI Search: being recognized as a source before being visited as a destination.
What makes content retrievable
Search Engine Journal lists several criteria. I confirm them across my deployments.
Retrievable content is content that AI engines can:
Extract easily (clear structure, clean semantic HTML tags, no JavaScript hiding text content)
Reuse without rewriting (raw data, sourced figures, precise definitions, tables)
Cite with confidence (recognized domain, inbound backlinks, recurring mentions in other cited content)
Example: a site I restructured in December 2025 (fintech, regulatory compliance). Before restructuring, the blog had 140 "général advice" articles. After: 52 technical pages.
Each page contains:
A sourced regulatory data table (Banque de France, AMF, ECB)
A precise definition of each technical term
An identified author (internal expert, LinkedIn linked)
A visible last-updated date
Outbound links to official sources
Results after 4 months:
Citations identified in ChatGPT Search: 28 occurrences
Organic traffic: -19%
Qualified incoming requests: +94%
The site didn't gain traffic. It gained perceived authority. AI engines cite it. Prospects arrive pre-convinced.
Dan Taylor (SEJ) sums it up: "Success now revolves around who gets surfaced, cited, and reused." You're not playing to be clicked anymore. You're playing to be cited.
What I'm concretely changing on client sites
Since November 2025, I apply a restructuring framework to every site I audit. Here are the 6 systematic changes:
1. Remove unsourced generic listicles Any article "Top 10 tips for X" without figures or sources = deleted or redirected. AI engines don't cite listicles. They cite raw data.
2. Add reusable data tables Every technical page gets at least one HTML table (not image, not PDF) with sourced data. Example: "Comparative EU VAT rates table 2026" with link to europa.eu.
3. Clear author identification Every article gets an author block with name, title, LinkedIn link. AI engines weight author authority in their citation scoring.
4. Primary source mentions Every figure = sourced. Every table = sourced. Every regulatory claim = official source linked. AI engines prioritize content citing sources they already cite.
5. Reinforced semantic markup schema.org Article + FAQPage standard. <time> tags for dates. <cite> tags for sources. AI engines parse HTML, not visual rendering.
6. Visible dated updates Every page gets a "Last updated: [date]" mention in HTML and visible rendering. AI engines weight freshness in their citation scoring.
Results observed across 12 restructured sites between November 2025 and February 2026:
Average traffic: -31%
Citations identified (Perplexity, ChatGPT, Bing Copilot): +340% (order of magnitude, manual count via targeted searches)
Qualified incoming requests: +112%
Traffic drops. Authority rises. Requests explode.
The flaw no one wants to see
Here's what strikes me most: agencies are still selling traffic.
I get competitor audits every week. All promise "+50% organic traffic in 6 months". None mention retrievability. None measure citations. None track perceived authority.
Result: clients paying to increase a metric that measures nothing.
An example: an e-commerce client (premium furniture, $3,800 average order) contacted me in January 2026. He'd been working with an agency for 8 months. The agency produced 120 blog articles. Traffic up +34%. Revenue down -11%.
I looked: the 120 articles are generic listicles ("15 decor ideas for your living room", "How to choose a couch"). Zero technical data. Zero tables. Zero sources. Google ranks them well. Perplexity never cites them.
Result: cold traffic, bouncing visitors, zero authority signals.
We stopped production. Restructured. Built 28 technical pages: care guides by material, European furniture standards, standard dimension tables, sourced solid wood spec sheets.
After 90 days:
Blog traffic: -27%
Perplexity citations identified: 19 occurrences
Qualified quote requests: +78%
Revenue: +23% (attributed partly to better-qualified incoming requests)
Traffic dropped. Revenue climbed.
The flaw is that everyone still measures what's easy to count (sessions, page views, duration) instead of what generates value (citations, perceived authority, qualified requests).
How to track citability
No unified tool yet automatically tracks all your citations across all AI engines. But you can measure manually.
Here's the method I've applied since November 2025:
1. Identify your target queries List the 20 technical queries where you want to be cited (not ranked — cited). Example fintech sector: "PEL rate 2026", "Livret A ceiling 2026", "tax bracket calculation 2026".
2. Query each AI engine manually Ask each query in Perplexity, ChatGPT Search, Bing Copilot, Google AI Overviews. Note if your site appears in cited sources.
3. Count monthly occurrences Retest each month. Track citation count evolution. If you move from 8 citations in January to 23 in March, you're gaining retrievability.
4. Cross with incoming requests Compare citation count with qualified requests. If both curves climb together, you've proven citability generates value.
I do this for 47 sites. Takes 2 hours per month per site. It's manual. But it gives crystal clarity on what works.
Quick alternative: set up Google Alerts on "[your brand] + Perplexity" and "[your brand] + ChatGPT". You'll get notifications when someone mentions seeing your site cited in an AI response.
DOSE angle applied: Measuring citability activates oxytocin (recognition as reliable source) and serotonin (perceived status in your sector). You're not playing for anonymous clicks anymore. You're playing to be publicly recognized as an authority.
What this changes for you
If you keep measuring SEO success only via Google Analytics, you'll panic. Your traffic will drop. Sessions will plummet. Page views will collapse.
But if you look at:
How many times you're cited in AI engines
How many qualified incoming requests you get
How many brand searches you capture
Your visitor-to-conversion rate (not volume)
You'll see that you haven't lost performance. You've gained authority.
Example HR SaaS client (180 employees, training management tool): traffic -34% between October 2025 and March 2026. Demo requests +141% same period. Visitor-to-demo conversion rate: +89%.
Translation: fewer visitors, but better-qualified ones who arrive already convinced because they saw the site cited in a Perplexity answer.
Dan Taylor (SEJ) states it clearly: "Content quality still matters, but it is no longer the deciding factor."
Quality no longer suffices. What matters is structure. Attribution. Sourcing. Retrievability.
You're no longer playing to write the best article. You're playing to write the most easily extractable, citable, reusable article.
That's a total mental shift. But that's what I observe on the ground after 6 months.
Retrievability audit: 47 queries tested live
First call = live audit of your site across 47 target queries in Perplexity, ChatGPT, Bing Copilot. I show you how many times you're cited, how many times you should be, and what's blocking it.
Query Perplexity, ChatGPT Search, Bing Copilot manually on your target queries. Note if your site appears in the cited sources. Repeat monthly to track evolution.
My traffic is down 30% but requests are up. Is that normal?
Yes. AI engines cite your content without redirecting users to your site. You lose traffic but gain perceived authority. Measure qualified requests, not sessions.
What is retrievable content concretely?
Content easily extracted by AI engines: sourced data tables, precise definitions, identified author, clean semantic markup, links to official sources.
Should I stop producing generic content?
If it contains no sourced figures, data, or reusable tables, yes. AI engines don't cite generic listicles. Prioritize technical sourced pages.
Does the DOSE framework apply to AI Search?
Yes. Oxytocin (recognition as a source) becomes the central lever. Being cited builds perceived trust, even without clicks. Taught by Guillaume Attias at BMO Academy.
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.
