Citation Share Replaces Impression Share: The 90-Day Playbook to Dominate AI Search
Summarize this article with AI
He was spending €45,000 a month on Google Ads. His traffic was eroding anyway.
A client calls me on a Tuesday morning.
Design furniture e-commerce site. €45,000 monthly in Google Ads. 4,200 organic sessions. A catalog of 1,200 SKUs.
And a drop in visibility over two quarters.
Despite maxed-out ads. Despite an agency charging €8,000 monthly for management.
« Stéphane, we’re losing qualified traffic. Search Console shows drops on our product pages. I don’t know why. »
I pull three long-tail queries on his domain. Zero appearances in Google AI Overviews. Almost nothing in Perplexity responses either.
I run a dozen buyer questions through ChatGPT: « which ergonomic wooden chair for remote work design »… « modular sofa comparison French-made furniture ».
His site is never cited.
Not even on searches where he’s had featured snippets for two years. Zero presence.
The error wasn’t the content. It was architecture. His domain was invisible to AI eyes because no semantic cluster linked his pages together to demonstrate expertise to answer engines. And all his Ads budget was bleeding away on increasingly expensive bids, never building citation share.
That’s where the work began.
Why impression share no longer holds weight against citation share
Jason Shafton worked inside Google Ads. His team grew the platform by billions of dollars. He knows the system closely. In his latest article for Search Engine Journal, he states a blunt truth: « You can’t bid to appear in an AI response. Six signals decide if assistants cite your brand, and most live off your site. »
He draws three lessons from years of bidding:
- Platforms reward the signals they measure, not your effort.
- Budget must follow evidence, not habit.
- Each tech shift advantages whoever adapts first.
Impression share was the star of the bidding years: what percentage of your ads showed against competition. This metric no longer exists for organic traffic in 2026. It doesn’t capture whether your pages appear in responses generated by language models. The real strategic lever is citation share: the share of AI responses that explicitly mention your domain as a source.
Take the query « best natural latex mattress reviews 2026 ». Before, your goal was position 1. Today, the AI Overview aggregates reviews directly in the SERP after a summary. The click no longer lands on your blue link, but on the source link embedded in the answer box. If your page isn’t cited, it doesn’t exist. Literally.
Citation share measures this presence. It works like a new Quality Score, but for large language models. The more structured your pages are as clear entities, the more they get cited. The less structured, the more they vanish.
What works well is that the signals these AIs measure — topical authority, semantic coherence, internal link depth — are exactly those my semantic clusters have been strengthening since 2016. No magic. Just architecture.
Phase 0: Measure your true citation share before spending another euro
Before touching a single Ads dollar, I forced a citation share audit on the furniture client.
Simple method, replicable in one morning:
- Select your top 50 historical transactional queries (Search Console + Ads data).
- Type each query on Google and extract AI Overview content if present.
- Note whether your domain appears in cited sources (sometimes the brand name is mentioned after the summary even without a direct link).
- Run the same queries via Perplexity or connect to the OpenAI API to simulate a user and collect citations systematically.
Result for this client: out of 45 queries where an AI Overview existed, his domain was cited on… 2. That’s 4.4% citation share, while he captured 22% impression share organic on those same queries and spent €18,000 monthly in Ads on those keywords.
22% down to 4.4%. A chasm.
This simple measurement let me give a diagnosis with confidence. The domain wasn’t structured to be cited. His product pages were isolated. His blog articles, even well-written, were never connected to his catalog through intentional entity linking. The AI didn’t understand « who was talking » about what.
That’s where the 90-day playbook makes sense.
Le client a enregistré trois changements majeurs : ses clics issus des AI Overviews ont explosé (+193%), sa citation share sur requêtes transactionnelles atteint 27%, et son budget publicitaire a pu être réduit de 62% sans perte de revenu. Voici le comparatif concret.
Avant/Après : Visibility & Cost Metrics
Les 3 KPIs clés transformés en 90 jours
Le playbook suit une logique stricte : diagnostic, cluster, activation, optimisation. Chaque phase construit sur la précédente. Ce flux montre comment passer d’un audit citation share nul à une domination sémantique en trois mois.
Architecture Sémantique : Le Framework 90j
Les 4 étapes séquentielles pour capturer la citation share
Day 1-30: Build cluster architecture before writing a single word
The error I see everywhere: produce more content to « feed the AIs ». Exactly backwards.
Jason Shafton echoes this too: « It’s not a volume competition, it’s a structured-evidence competition. »
I set one rule with this client: we stop. No more articles drafted, no product updates without entity logic. We go back to source: how should the domain be carved so each catalog segment becomes a strong entity?
We mapped:
- 6 main clusters: « design chairs », « modular sofas », « artisan tables », « lighting », « custom furniture », « space planning advice ».
- Each cluster contains a pillar page (the page that defines the entity), satellite pages (each SKU with links back to content pillar), and connective content for each secondary entity/parent entity pair.
In 30 days, the team created 47 pillar pages and reworked the structure of 294 existing pages. Nothing more.
We reworked structured data to clarify entities (Organization, Product, FAQ, Article, entity mentions in descriptions). We strengthened internal linking, jumping from 1.2 average links between cluster pages to 4.7 links.
Total cost: €12,000 consulting + internal writing time. Zero Ads budget cut. Just architecture.
By day 30, citation share on the top 45 queries already climbed to 9.8%. Without spending a cent on feeding content to AIs.
The key insight, Days 1-30
If your site isn’t a web of clear entities, language models won’t cite you. They hunt for sources that actually know what they’re talking about. A semantic cluster proves it.
Day 31-60: Activate signals and shift citation share
Once the foundation was in place, we triggered two things:
1. Produce the missing content—only what the cluster demands.
Each pillar page was tied to a gallery of satellite pages forming a semantic cluster. Where gaps existed (comparator queries, specific buyer questions), we filled them with surgical production: 93 satellite pages drafted in 30 days by a pool of writers trained in DOSE format. Non-negotiable: each satellite links to two other satellites in the cluster and the pillar page, and mentions complementary entities at least once.
2. Install a citation share tracking dashboard.
You can’t drive without visibility. With help from a Google Sheets script plus the Google Custom Search API, we automated source monitoring for AI Overviews on the top 100 queries. Every Friday, the script outputs a citation share report by cluster and entity.
The curves are brutal when you see them: on the « modular sofas » cluster, citation share jumped from 2.1% to 14.7% in 20 days, just from better linking. AIs started treating the domain as an authority.
This phase cost €6,500 (writing + tool setup), and in parallel we began cutting some Ads bids on queries where citation share climbed, following Shafton’s principle: budget follows evidence.
Result day 60: 16.4% citation share on the top 45 queries. Organic traffic grew from 4,200 to 7,800 sessions. And Ads costs dropped to €38,000 (first a cautious €7,000 cut, no impact on conversions).
Voici le chemin exact suivi par le client meuble : partant d’une répartition 100% Ads, les 90 jours de clustering architectural et d’activation sémantique ont progressivement décalé le budget vers l’organique, sans perte de revenu. Chaque étape montre les gains de citation share et la réallocation correspondante.
Budget Reallocation & Citation Share Growth
Comment 90 jours transforment la répartition budget et capturent 27% de citation share
Day 61-90: Optimize for citations and reallocate Ads budget
The last 30 days: that’s where the portfolio truly shifts.
On one side, we continue thickening cluster linking: each satellite page gets additional contextual links, we add structured FAQs on pillars, and we create « comparison » pages in clusters. The domain asserts itself on target entities. Citation share rises to 27.1% by day 85.
On the other, we apply the bidding lesson: budget follows evidence, not habit.
I ran the client through an Ads bid review by cluster. For each query, we looked at citation share, organic post-click conversion rate, and Ads cost per conversion. Once a query crossed 15% citation share with stable organic conversion, we gradually lowered the Ads bid.
On the « artisan tables » cluster, Ads spend dropped from €8,200 to €2,900 in 20 days. Revenue didn’t move. The share of sales driven by organic traffic through AI Overviews jumped from 12% to 34%.
Result at day 90:
- Citation share on top 50 queries: 27.3%
- Organic sessions: 12,200 (vs 4,200 initially), of which 3,800 from AI Overviews and AI citations
- Monthly Ads budget: €17,100 (-62%)
- Revenue driven by organic share: +€23,000 monthly
You might think AI Search would kill traffic. Actually, it shifted it toward those with readable architecture.
This client no longer pays to buy visibility. He is the visibility.
Why your semantic clusters are your best bet in 2026
Jason Shafton pushes one point I validate 100%: signals live on your own domain. AI assistants don’t rely on random backlinks. They extract information from sources that actually cover a topic end-to-end. That’s the topical authority my semantic clusters build.
Without a cluster, each catalog page fights alone. With a cluster, your whole domain shouts: « I’m the expert on this subject, my pages prove it by linking to each other. » That’s exactly what language models hunt for.
In 90 days, this client built 6 complete clusters, 47 pillar pages, 93 satellites, linking that climbed from 1.2 to 4.7 average links intra-group. He cut Ads spend by 62% while growing organic traffic +190%. Citation share quantifies that gain.
You don’t need a site redesign. You need an entity schema.
What if you ran this playbook on your own domain?
I review 15 sites a week. Most burn fortunes on Ads for queries they could dominate with citation share in 4 weeks of architecture.
The diagnosis always returns to the same truth: isolated pages, frozen Ads budgets, zero visibility in AI Overviews.
I’m not selling magic software. I’m showing you your pages. The ones that could be cited if they were connected. The ones missing.
In one call, I scan your domain through the lens of AI citations and tell you which clusters to activate first. No theory. Evidence.
You now know how Google’s former Ads director would do it. The only question left: in 90 days, will the share of AI responses carrying your brand be 4% or 27%?
Your citation share in 20 minutes
I scan your domain on 40 queries, identify your citable pages, and give you the list of clusters to activate first. No pitch. Just your numbers.
Book a strategic call — 45 minFrequently Asked Questions
How do I measure my citation share without complex scripts?
Select 20 to 50 strategic queries. Search them on Google logged out and note how many times your domain appears in the AI Overview or in the « people also ask » box sources. Average it out. Even manual, this method gives you a reliable citation share indicator.
How fast can citation share grow?
<p>With my clients, first improvements land between day 30-45, once internal linking works without even new content. It accelerates around day 60 when satellites are connected and AIs recalculate topical authority.</p>
Should I kill my Google Ads budget to bet on AI Search?
No. The playbook is to reduce it progressively as citation share climbs. Keep Ads on highly intentional queries that haven’t yet scored AI citations. The budget you free from wasted Ads, you reinvest in cluster building and tracking tooling.
Are semantic clusters enough to guarantee a spot in AI Overviews?
They’re the necessary condition. Without entity architecture, your pages stay isolated and AI judges them unreliable. Add correct structured data and content answering real buyer questions: your odds improve sharply.
Which queries should I target first to win fast on citation share?
Target « comparison » and « best [product] for [use case] » queries where you can showcase what you do and many advertisers bet on paid ads. These are where your ad impression share and citation share diverge most — where you’ll gain fastest.
