ChatGPT Changed Their Algorithm 46 Days Before Announcing Ads

After ChatGPT's January 16 ad announcement, we looked at the data. What we found was remarkable: the algorithm had been changing since December 1. This is the forensic investigation.

On January 16, 2026, ChatGPT announced ads. A week later, we noticed something odd in our LLM visibility data. When we investigated, we uncovered a smoking gun: the algorithm had been quietly changing for 46 days before the public announcement. This is the story of what we found when we looked back at the data.

The Investigation: What Made Us Look

What We Noticed (Late January 2026)

After ChatGPT announced ads on January 16, we started seeing unusual patterns in our client data. Brand visibility was behaving differently. Citation patterns looked strange. The post-announcement data didn't match what we expected. So we decided to investigate: had the algorithm already been changing before they told anyone?

We went back through our 206,412 tracked ChatGPT responses across 43 brands. We analyzed November, December, and January data looking for algorithmic shifts. What we discovered was startling: the changes began in early December—46 days before the public announcement.

The Finding

ChatGPT didn't just announce ads and then change their algorithm. They changed the algorithm first—starting December 1, 2025—and announced ads 46 days later. Our forensic analysis of the data reveals the full timeline of preparation.

The Timeline: What the Data Revealed

November 14-30, 2025 - Baseline Period

Normal operations. Visibility rate: 41.3%. Citations per response: 5.7. This was our control period—ChatGPT operating as users expected. No signs of change yet.

~December 1, 2025 - First Signal: Citation Engine Activated

Looking back at the data, we found citations per response jumped 81% (5.7 → 10.4) starting December 1. At the time, we didn't know why. Now we understand: they were building credibility infrastructure for ads.

December 2025 - Testing Phase (Discovered Retroactively)

Our forensic analysis revealed visibility increased to 46.2% (+4.9pp) throughout December. Stage distribution shifted toward commercial intent. "Advice" queries grew 31%. They were testing the algorithm in production—silently.

January 1-15, 2026 - Re-Ranking Deployed (Before Announcement)

The data shows visibility normalized to 43.7% in early January. Brand rankings shuffled. 55.8% of brands lost visibility. This created "scarcity"—all before the public knew ads were coming.

January 16, 2026 - Public Ad Announcement

ChatGPT officially announced ads. This is when the public learned what was coming. But the algorithm had already been changing for 46 days.

~January 23, 2026 - Our Investigation Begins

We noticed odd patterns in post-announcement data and decided to look back. Our forensic analysis of 206,412 responses revealed the full 46-day timeline of algorithmic preparation.

Post-Investigation

Visibility continues at 37.3%. Algorithm stable. Now we understand: ChatGPT spent 6+ weeks preparing the monetization infrastructure before the public announcement. This is the new platform playbook.

What the Forensic Data Shows

Statistical evidence from 206,412 ChatGPT responses across 43 brands:

• 46 Days: Algorithm modified before public disclosure
• +81% Citation Increase: 5.7 → 10.4 citations per response (Dec 1 onward)
• +31% Advice Stage Growth: High commercial-intent queries prioritized
• 55.8% Brands Impacted: Lost visibility before announcement
• Statistical Significance: p<0.01, all findings statistically validated
• Effect Size (Cohen's d): 0.67-0.75, medium-to-large practical effects
  
  

Five Pieces of Evidence

Our retroactive analysis uncovered five distinct signals of algorithmic preparation—all occurring before the January 16 announcement.

1. Algorithm Changed 46 Days Early

The Smoking Gun: When we analyzed historical data, major shifts appeared starting December 1, 2025—a full 46 days before ChatGPT publicly announced ads.

• Citations: 5.7 → 10.4 (+81%)
• Timeline: December 1, not January 16

What This Means: ChatGPT spent 6+ weeks preparing the algorithm infrastructure in secret. The announcement came after the changes were already deployed.

2. Commercial Intent Prioritization

What We Found: High-commercial-intent query stages grew significantly throughout December and early January (p=0.010):

• Advice stage: +2.1pp (31% relative increase)
• Evaluation stage: +5.5pp (71% relative increase)
• "Other" (low-intent): -10.9pp decrease

Why It Matters: These are the queries advertisers pay premium rates for. ChatGPT was recalibrating for monetization before telling anyone.

3. Pre-Announcement Visibility Re-Ranking

The Pattern: Brand visibility changed dramatically in December and early January—before the ad announcement.

• Mean visibility decline: -4.2pp
• Brands losing visibility: 55.8%
• Statistical significance: p=0.009

The Implication: They created organic reach "scarcity" before launching the paid product. Classic platform monetization playbook.

4. Citation Source Strategy Shift

What Changed: Citation patterns shifted toward platforms with existing ad infrastructure:

• Reddit: +5.2% growth (went public March 2024, has ads)
• Medium: 3,326 new citations (subscription model)
• Social/commercial platforms: +7.0pp overall increase

The Signal: ChatGPT was testing partnerships with monetizable platforms—before the public knew ads were coming.

5. Announcement Day Anomaly

The Spike: On January 16 (announcement day), visibility jumped to 45.98%—then immediately returned to the declining trend.

• Pre-announcement (5-day avg): 37.3%
• Announcement day: 45.98% (+8.7pp spike)
• Post-announcement: 37.3% (back to trend)

Interpretation: Possible "showcase" effect to minimize backlash on announcement day, then returning to the new algorithm baseline.

The Bottom Line: Platform Changes Happen Before Announcements

We didn't predict ChatGPT would introduce ads. But when they announced it, we had the data to investigate—and we discovered the algorithm had been changing for 46 days before the public knew.

This is the new reality: platforms modify algorithms in secret, announce changes publicly later, and most brands don't have the visibility data to understand what happened. We do.

What This Means for Brands

The New Reality

• Platforms change algorithms before announcing—you can't rely on public disclosure timelines
• By the time ads are announced, the game has changed—organic visibility has already been reduced
• Historical data analysis is crucial—you need to be able to look back and understand what happened
• Continuous monitoring is essential—you can't wait for announcements to react

Strategic Implications

• Assume ChatGPT ads are fully operational—the algorithm has been ready since early January
• Other platforms will follow this playbook—Perplexity, Claude, Gemini may already be changing
• LLM visibility tracking is non-negotiable—you need portfolio-wide data to see platform-level patterns
• Forensic analysis capabilities matter—can you look back 60-90 days when something seems off?

Learn more about tracking and optimizing for AI platforms with our comprehensive GEO research..

What Worked for This Investigation

• Continuous historical tracking: We had 69 days of data to analyze retroactively
• Portfolio-wide visibility: 43 brands showed us patterns individual brands couldn't see
• Statistical rigor: We could prove changes were real, not noise (p<0.01)
• Multi-metric analysis: Visibility + citations + stage distribution told the full story

Don't Get Caught Without Data

When ChatGPT announced ads, we had the historical data to investigate. We discovered they'd been changing the algorithm for 46 days. What's changing right now that you can't see?

Scrunch AI + Seer's LLM visibility tracking gives you the data you need—before and after announcements. Let's talk.

About This Research

Forensic analysis conducted by Seer Interactive's Research Lab using Scrunch AI LLM visibility tracking platform.

© 2026 Seer Interactive. All findings and data are proprietary.