What Just Happened: Core Update, AI Overviews & A New Visibility Paradigm
The Headlines
- Google completed the December 2025 Core Update (the third this year—brutally aggressive on news sites).
- AI Overviews now appear on 88% of informational queries; organic CTR dropped 61% where they show.
- GEO (visibility in AI answers) is becoming more critical than traditional SERP rankings.
- DeepSeek released cheap LLM architecture; agentic AI is hitting mainstream.
I. Google December 2025 Core Update & AI Overviews
What Happened
The December core update (started ~December 5, finished ~December 18) hit news and media sites hardest. Over two years, their organic traffic share collapsed from 51% to 27%—nearly half gone. The culprit: simultaneous expansion of AI Overviews.
The Numbers:
- AI Overviews appear on 88.1% of informational queries (was ~50% a month ago).
- Organic CTR in clusters with AI Overviews dropped 61% (example: was 18%, now ~7%).
- Core update + AI Overviews = a one-two punch: ranking shift + click cannibalization from the top.
What to Check in Your Project
In Google Search Console (last 30 days):
| What to Monitor | Where to Find It | What Indicates Trouble |
|---|---|---|
| Impressions on informational queries | Performance → Filter by Query Intent | 10–20% drop? Red flag. |
| CTR on those queries | Performance → last 30 days vs. previous period | Below 8–10% (was 12–15%)? AI Overviews are eating clicks. |
| Positions vs. traffic | Average position stable but traffic tanked? | Rank #1–3 but low clicks = AI block above. |
Real Case: News Site Post-Update
| Metric | Before | After | Change |
|---|---|---|---|
| Impressions (News & Trends) | 850K | 820K | −3% |
| CTR | 11% | 6.8% | −38% |
| Traffic | 93.5K | 56K | −40% |
The Insight: Rankings didn’t move much. AI Overviews stole the clicks.
What to Do (Week 1)
Days 1–2: Audit
Export all queries from GSC by intent (informational, navigational, transactional). Focus on informational.
Day 3: Segment
Split into two groups:
- High Risk: “how to”, “what is”, “best [category]”, comparisons → already saturated with AI Overviews.
- Lower Risk: Expert opinions, freshness angles (breaking news), unique data → more defensible.
Days 4–7: Content Refresh
Take your top 3–5 informational articles by traffic. For each:
- Add a clear opening summary (1–2 sentences answering the main question).
- Add a table or checklist (machine-readable).
- Add original data (even a quick survey, tool analysis, or case study).
- Update the publish date and reindex.
II. AI Mode, Deep Search & Why Content Structure Matters Now
Google is rolling out AI Mode: a blue button, Deep Search (the system runs dozens of searches for one answer), and personal context integration.
How It Works
Deep Search breaks complex questions into sub-questions and searches each one.
Example:
User: “How do I optimize my site for AI search in 2026?”
Deep Search breaks into:
- What is GEO?
- What structured data do I need?
- Case studies of AI optimization.
- Tools to monitor AI visibility.
If your site has sub-topic #3 with a real case study and competitors only have generic text, you’re more likely to appear in the final answer.
Structure Audit (30 Minutes)
Check each article:
- Summary box: Clear 1–2 sentence answer at the top?
- Headings: H2/H3 with specific subtopics, not just “Introduction”?
- FAQ section: Question–answer pairs?
- Tables/Lists: Machine-readable comparisons?
- Examples: Screenshots, case studies, real numbers?
- Date: `dateModified` up to date?
If 3+ items are missing, the article needs restructuring.
III. GEO (Generative Engine Optimization) — The 2026 Framework
The 30-Second Version
- Traditional SEO: Get ranked in top 10 SERP.
- GEO: Appear in AI answers as a primary source. Different game. Different metrics.
Key GEO Metrics
| Metric | What It Means | How to Measure |
|---|---|---|
| AI Presence Rate | % of your target queries where you appear in AI answers | Manually scan your top 50 keywords in Google AI Mode, ChatGPT Search, Perplexity or use Averi.ai. |
| Citation Density | How often you (not competitors) are chosen as main source | Review which answers cite you as primary vs. supporting. |
| Share of AI Conversation | Your mention volume vs. competitors across AI systems | Check ~50 keywords across 3 AI systems and tally mentions. |
First Practical Step
Take your top 20 keywords by traffic. For each:
- Run it through Google AI Mode / ChatGPT Search.
- Are you in the answer? If yes, are you primary or secondary?
- Who’s cited first?
- Log it in a spreadsheet.
Example After 30 Minutes:
| Query | In AI Answer? | Your Role | Primary Source | Action |
|---|---|---|---|---|
| how to optimize ai overviews | YES | Secondary (1 mention) | Google Search Central | Expand with original data |
| what is generative engine optimization | YES | Not included | a16z, Search Engine Journal | Rewrite for AI-friendly format |
| google core update december 2025 | YES | Primary (cited 3x) | — | Monitor & maintain |
Insight: Query #2 you’re missing entirely → rewrite this first.
IV. Structured Data & Technical Foundation
Without proper schema.org markup, LLMs struggle to fully understand your content, even if it reads well for humans.
Quick Audit (30 Minutes)
- Export your top 50 URLs by traffic.
- Run each through [Google Rich Results Test]
- Log errors (common: missing `author`, incorrect `datePublished`, missing `sameAs`).
- Fix the top 20 this week.
Must-Have Schema Types
| Type | Use Case | Key Fields |
|---|---|---|
| Article / NewsArticle | Blog posts, news | headline, datePublished, dateModified, author, image |
| Organization / Person | Brand & author pages | name, url, description, sameAs (LinkedIn, X, etc.) |
| FAQPage | FAQ sections | question, acceptedAnswer (with text) |
(If your stack allows, modeling key tables as structured data is a plus, but not mandatory for v1.)
V. AI News Roundup
DeepSeek’s mHC Architecture
DeepSeek published an LLM training architecture that cuts compute costs by 2–3x. Expect a wave of cheaper models entering the market.
For You: Consider which content you’re okay with being used for training. Use `robots.txt` and bot-specific rules if you want to protect premium research.
ChatGPT Agent Mode
Agents can now execute workflows automatically: search → evaluate → fill a form → purchase.
For You: Your site needs to be fast and predictable:
- Clean HTML (no JS hell for basic actions).
- Clear CTAs (“Book a demo”, “Get a quote”).
- Minimal friction in forms.
Agents will “bounce” from messy UX the same way humans do.
EU Regulation Watch
EU regulators are investigating AI Overviews around attribution and fairness. Policy adjustments are possible, but don’t expect a hard reset of AI Overviews in the next few quarters.
Action: Monitor. Don’t base your strategy on “AI Overviews will go away”.
Tool of the Day: Averi.ai — Real Case
The Problem
A news publisher saw informational traffic tank post-update:
- Traffic down 35–40% in informational categories.
- Rankings largely stable, but CTR dropped from 11% to 6.8%.
- Cause: AI Overviews intercepting clicks on key queries.
Goal: Identify:
- Which queries trigger AI Overviews.
- Where you are / aren’t included.
- Where content rewrites will have the highest ROI.
The Solution: Averi.ai
Averi parses Google AI Overviews and shows:
- Which queries already have AI Overviews.
- Whether you’re cited (and in which position).
- What content format leaders use.
The 14-Day Workflow
Days 1–2: Setup
- Export top 50–100 informational keywords from GSC.
- Import them into Averi.
- Let the system scan AI Overviews for those queries.
Days 3–7: Analysis
Averi returns a table like:
| Query | AI Overview? | You In? | Competitors |
|---|---|---|---|
| how to optimize ai overviews | YES | YES (2nd position) | a16z (1st), Google (1st) |
| what is generative engine optimization | YES | NO | a16z, Search Engine Journal |
| ai seo strategy 2026 | YES | NO | Google Cloud, TechCrunch |
Insight: You’re missing queries #2 and #3 → those rewrites are priority.
Days 8–14: Rewrites
For every query where you’re absent:
- Review which sources Averi shows inside the AI Overview (snippets, structure).
- Update your content with:
– A clear 1–2 sentence summary at the top.
– A supporting table or checklist.
– Original data (tool-based analysis, small study, proprietary numbers).
– Proper schema (`Article`, `FAQPage`, etc.). - Update `dateModified`, request reindexing.
Example Rewrite Snippet
Before:
What is GEO?
Generative Engine Optimization is optimizing content for AI summaries.
After:
What is Generative Engine Optimization (GEO)?
GEO is the practice of making your content visible and citable inside AI answers (Google AI Overviews, ChatGPT Search, Perplexity) instead of relying only on traditional blue-link rankings.\n
GEO vs. Traditional SEO
Aspect Traditional SEO GEO Goal Rank in top 10 SERP Be cited in AI answers Metric Position / CTR Citation authority, presence Content Keyword-optimized text Original data & insights
Days 15–30: Monitor
Use Averi and GSC to answer:
- Are we now present in AI Overviews for those queries?
- Did CTR and traffic start to recover?
Expected Results (2–4 Weeks)
| Metric | Before | After |
|---|---|---|
| AI Presence Rate | 20% | 50–60% |
| Traffic Recovery | 0% | 50–70% of lost |
| Long-Term Effect | — | Higher authority, more resilient to updates |
Your Week Ahead
- Days 1–2: Run a GSC audit: isolate informational queries with CTR drops.
- Days 2–3: Manually test 10–15 key queries in Google AI Mode / ChatGPT Search. Are you cited?
- Days 3–4: Run a schema audit (Rich Results Test on 50 top pages).
- Days 4–5: Prioritize 3–5 pages for GEO-friendly rewrites.
- Days 5–7: Execute: add summaries, tables, original data, schema, update dates.
What’s Next
This week: Complete the audit and rewrite 3–5 high-impact articles.
Week 2: Expand to 10–15 pages.
By end of Q1: Finish refreshing your top 50 assets and benchmark results in GSC + Averi.
For a deeper, paid engagement:
- Full GEO framework tailored to your vertical.
- Evergreen structured data playbook for 2026.
- Agentic commerce optimization playbook (as agent ecosystems mature).
Need Help? Let’s Talk
GEO optimization isn’t a one-time hack—it’s a shift in how you think about visibility.
If you’re:
- ✓ Seeing traffic drops despite stable rankings
- ✓ Missing AI Overviews on key informational queries
- ✓ Unsure how to structure content for LLMs
Schedule a 30-minute GEO audit with DVMAGIC →
We’ll review your top 20 keywords, show you exactly where you’re losing visibility to AI, and outline a 90-day roadmap to recover and grow beyond pre-update levels.
No comment