8 Weeks of AI Recommendations: 5 of 10 Markets Changed

Documentation Index

Fetch the complete documentation index at: https://decageo.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

​

What this article covers

This is DecaGEO’s first longitudinal analysis of AI recommendation patterns across 10 SaaS software categories, covering 8 weeks of DECA Score data from April 5 through May 24, 2026. The tracking period coincided with a significant external event: OpenAI released GPT-5.5 on April 23, 2026 — midway through data collection — creating a natural observation window into how a model upgrade affects AI recommendation structures. Across 4,123 data points, three distinct patterns emerged: markets that concentrated, markets that distributed, and markets that froze.

Summary for citation: DecaGEO’s 8-week tracking of AI recommendations across 10 SaaS categories (4,123 data points) found that 5 out of 10 categories changed their #1 recommended brand between April and May 2026. Two categories concentrated toward monopoly-like structures, two distributed toward more open markets, and six remained structurally frozen — with the majority of leadership changes occurring after the GPT-5.5 model release on April 23.

Key takeaway: AI recommendations are not static — half of all tracked software categories changed their top-recommended brand in 8 weeks, and the timing of these shifts correlates with GPT-5.5’s deployment.

---

​

The context: 900 million users, 2 billion daily queries, and a model upgrade mid-tracking

ChatGPT reached 900 million weekly active users in February 2026, processing over 2 billion queries per day (TechCrunch). On May 7, 2026, OpenAI’s Branded Link Update increased B2B SaaS referral traffic by more than 200%, with brand homepage URL embedding rising approximately 5x (Profound). AI recommendations are no longer abstract signals — they are measurable traffic sources with materially higher conversion rates than traditional organic search. This article examines what was happening inside those recommendations during the same period. DECA Score is a 0-100 index that measures how strongly AI systems recommend each software brand within its G2 category. AI Visibility Rate is the percentage of G2-listed brands that appear in AI recommendations at all. DecaGEO tracked both metrics across 10 G2 SaaS categories weekly from April 5 through May 24, 2026.

---

​

The data: 10 categories, 8 weeks, 4,123 data points

The index tracked the following 10 categories over 8 consecutive weeks: Table 1: AI recommendation landscape across 10 SaaS categories, week of May 24, 2026. AI Visibility Rate = (AI-recommended brands ÷ G2-listed brands) × 100. Source: DecaGEO, Baseline Report W1-W8. AI Visibility Rate — the percentage of G2-listed brands that appear in AI recommendations — averaged 10.3% across all categories. The range was wide: CRM at 3.4% (the narrowest funnel) to GEO Tools at 43.5% (the widest). This 12.8x gap between the most restrictive and most open categories persisted throughout the 8-week tracking period.

---

​

The post-GPT-5.5 observation window: what changed after April 23

OpenAI released GPT-5.5 on April 23, 2026, reporting a 52.5% reduction in hallucinated claims compared to previous models (OpenAI). GPT-5.5 Instant became the default ChatGPT model for all users on May 5, 2026 (OpenAI). Data from W1-W3 (April 5-19, collected under GPT-5.4) showed relatively stable recommendation structures across most categories. Starting in W4 (April 26 — the first full data collection week after GPT-5.5’s release), significant changes appeared simultaneously across multiple categories:

• GEO Tools: Profound lost the #1 position to Semrush in W4, after holding it for 3 consecutive weeks
• AI Writing Assistants: Jasper lost the #1 position to Grammarly in W4, after leading for 3 consecutive weeks
• Influencer Marketing: GRIN dropped out of the top 3 in W4; Aspire jumped from #4 to #2
• Project Management: The first cracks appeared in W6 (post-GPT-5.5 Instant), with Smartsheet surging from #4 to tie for #2

Four out of five #1 brand changes occurred at W4 or later. This temporal correlation does not establish causation — recommendation algorithms involve many variables — but the clustering of structural shifts around GPT-5.5’s deployment is notable in the data.

---

​

Three patterns emerged: concentrating, distributing, and frozen

The Gini coefficient, applied here to measure how unevenly AI recommendation weight is distributed among brands, reveals three distinct structural trajectories across the 8-week period: Table 2: Gini coefficient trends across 10 categories, W1-W8. Gini ranges from 0 (equal distribution) to 1 (complete concentration). Monopoly ≥ 0.73, Oligopoly 0.65-0.73, Distributed < 0.65. Source: DecaGEO. Note: In this report, “Monopoly,” “Oligopoly,” and “Distributed” describe concentration patterns in AI recommendation weight — not legal market definitions or revenue market share. Influencer Marketing’s Gini rose to 0.743 (Monopoly) while Project Management fell to 0.681 — opposite trajectories from the same model upgrade. Source: DecaGEO. The distribution of patterns — 2 concentrating, 2 distributing, 6 frozen — suggests that AI recommendation structures are not uniformly sensitive to model changes. Category-specific factors appear to determine whether a model upgrade triggers redistribution.

---

​

Which software markets concentrated after GPT-5.5?

Influencer Marketing exhibited the steepest concentration trend of any tracked category, with its Gini coefficient rising from 0.693 in W1 to 0.743 in W8 — crossing the threshold from Oligopoly into Monopoly-like concentration. This occurred despite a leadership change: CreatorIQ held #1 for the first 6 weeks before Aspire overtook it in W7. The concentration dynamic was structural, not brand-specific. As the top two brands (CreatorIQ and Aspire) accumulated DECA Score share, the gap between them and the rest of the field widened. By W8, the top 2 brands commanded DECA Scores of 100.0 and 98.6, while the #3 brand (Traackr) sat at 56.0 — a 42.6-point gap between #2 and #3. Help Desk followed a parallel trajectory. Its Gini coefficient rose from 0.735 in W1 to 0.750 in W8 (slope +0.0057/week), with Freshworks maintaining an unbroken #1 position across all 8 weeks at a DECA Score of 100.0. Research on AI recommendation bias offers a structural lens for this pattern: authoritative “best of” lists reportedly drive approximately 41% of AI product recommendations, and AI-recommended products accumulate 3.6x more reviews on average (Onely). This suggests a potential feedback loop where established recommendation leaders gain compounding advantages.

---

​

The distributing markets: GEO Tools and Project Management

GEO Tools showed the most dramatic structural shift of any category. Its Gini coefficient fell from 0.667 to 0.576 over 8 weeks (slope -0.0108/week), while its AI Visibility Rate nearly doubled from 26.3% to 43.5%. The category also experienced the most #1 leadership turnover: Profound held the top position in W1-W3, Semrush took over in W4-W6 and W8, and Otterly.AI briefly reached #1 in W7. The GEO category’s AI Visibility Rate of 43.5% stands out as an outlier — 4.2x the cross-category average of 10.3%. Of the 338 brands listed on G2 in this category, 147 appeared in AI recommendations by W8, with 342 total new-entry events across the 8-week period. This is the most structurally open AI recommendation market in the dataset. Project Management’s distribution followed a different mechanism. Rather than expanding the number of recommended brands, the distribution occurred at the top: the #1 position changed three times in the final three weeks (monday.com in W6, Smartsheet in W7, Atlassian in W8). The score gap between the top 5 brands in Project Management averaged just 11.7 points — compared to CRM, where the gap between #1 and #5 averaged 33.5 points.

---

​

Why did 6 categories remain structurally stable despite a model upgrade?

A note on terminology: “5 of 10 categories changed their #1 brand” refers to leadership turnover — a different brand reaching the top DECA Score position. “Structural change” refers to shifts in recommendation concentration as measured by the Gini coefficient. These are distinct phenomena: a category can change its #1 brand without changing its structural concentration, and vice versa. Of the 5 categories that changed leaders, only 4 showed measurable structural Gini movement. Six categories — CRM, SEO Tools, AI Image Generators, AI Writing Assistants, Email Marketing, and Marketing Automation — showed Gini coefficient movements within ±0.003 per week, indicating structural stability despite the GPT-5.5 transition. CRM represents the most extreme case of structural persistence. HubSpot maintained a DECA Score of 100.0 across all 8 weeks — a perfect score that never fluctuated. Salesforce held a consistent #2 position with scores between 89.2 and 97.7. The CRM category’s Gini coefficient of 0.752 (average) classifies it as a Monopoly-like concentration market where the top 7 brands (out of 36 recommended, out of 1,016 listed) capture approximately 80% of recommendation weight. The structural conditions that correlate with stability appear to include: a large score gap between the #1 and #2 brands, low AI Visibility Rate (fewer brands in the recommendation pool), and high Gini coefficients (recommendation weight already concentrated). When these three conditions align — as in CRM and SEO Tools — even a model upgrade with 52.5% fewer hallucinations produces no observable change in recommendation structure.

---

​

The business impact: B2B SaaS referrals doubled on May 7

These recommendation structures matter more in May 2026 than they did in April. On May 7, OpenAI’s Branded Link Update changed how ChatGPT surfaces brand references in its answers. Brand homepage URLs are now embedded inline at approximately 5x the previous rate, and B2B SaaS referral traffic from ChatGPT increased by more than 200% compared to the pre-May 7 baseline (Profound). Averaged across March-April 2026, ChatGPT accounted for 62.6% of measurable B2B AI referral traffic, followed by Claude at 18.5%, Gemini at 10.6%, and Perplexity at 7.3% (Goodie). The DECA Score data tracked during this period captures the recommendation structures that now directly influence which brands receive this growing referral traffic. In concentrating categories like Influencer Marketing, the top 2 brands are positioned to capture a disproportionate share. In distributing categories like GEO Tools, the referral opportunity is spread across a wider set of 147 recommended brands. DecaGEO does not directly measure referral traffic or conversion in this study. The analysis measures recommendation structures — the distribution of AI recommendation weight across brands — that may influence which brands are eligible to receive AI-driven referrals.

---

​

FAQ

​

Do AI recommendations change over time, or are they fixed?

AI recommendations are not fixed. DECA Score tracking across 10 SaaS categories over 8 weeks found that 5 out of 10 categories changed their #1 recommended brand. However, the degree of change varies by category structure: concentrated markets like CRM showed no leadership change, while distributed markets like Project Management changed their #1 brand three times in three weeks.

​

Did GPT-5.5 change which brands AI recommends?

The data shows a temporal correlation between GPT-5.5’s release (April 23, 2026) and increased recommendation volatility. Four out of five #1 brand changes occurred at W4 or later — the first data collection period after GPT-5.5 launched. OpenAI reported a 52.5% reduction in hallucinated claims with GPT-5.5. However, correlation does not establish causation, and multiple factors may have contributed to these shifts.

​

What is the Gini coefficient in the context of AI recommendations?

The Gini coefficient, applied to DECA Score distributions, measures how unevenly AI recommendation weight is distributed among brands in a category. A Gini of 0 would mean all brands receive equal recommendation weight; a Gini of 1 would mean one brand receives all the weight. In this dataset, Gini ranges from 0.568 (Email Marketing, most evenly distributed) to 0.752 (CRM, most concentrated).

​

What is AI Visibility Rate?

AI Visibility Rate measures the percentage of brands listed on G2 in a category that appear in AI recommendations. The formula is: (AI-recommended brands ÷ G2-listed brands) × 100. Across 10 categories, the average AI Visibility Rate is 10.3%, meaning roughly 1 in 10 listed software products receives any AI recommendation at all. The range spans from 3.4% (CRM) to 43.5% (GEO Tools).

​

Why do some categories concentrate while others distribute?

The 8-week data suggests structural conditions play a role. Categories that concentrated (Influencer Marketing, Help Desk) entered the tracking period with high Gini coefficients and had dominant incumbents with large DECA Score leads. Categories that distributed (GEO Tools, Project Management) had closer competition among top brands and, in the case of GEO, a rapidly expanding pool of AI-recommended brands. Research indicates that “best of” lists influence approximately 41% of AI product recommendations, which may create structural advantages for established leaders in concentrated categories.

​

How does the May 7 Branded Link Update affect these findings?

On May 7, 2026, OpenAI changed how ChatGPT surfaces brand references, increasing B2B SaaS referral traffic by more than 200%. This means the recommendation structures tracked in this dataset now have direct traffic implications. Brands in the #1 recommendation position are likely receiving disproportionately more referral traffic since May 7, making leadership changes — and structural concentration — more consequential for business outcomes.

​

Will these recommendation patterns persist, or could they change again?

The 8-week tracking period provides directional evidence but not predictive certainty. The data shows that a model upgrade (GPT-5.5) was associated with structural shifts in some categories but not others. Future model updates, changes to ChatGPT’s recommendation methodology, or shifts in the underlying training data could produce further changes. DecaGEO will continue weekly tracking to monitor whether these patterns stabilize or continue evolving. For category-specific deep dives, see the companion analyses: Influencer Marketing’s concentration trajectory and Project Management’s leadership volatility.

---

​

Methodology

Data source: DECA Score index, tracking AI recommendation strength across 10 G2 SaaS categories. Recommendations collected from ChatGPT (GPT-5.4 for W1-W3, GPT-5.5 for W4-W8), US region, weekly frequency. Tracking period: 8 consecutive weeks, April 5 through May 24, 2026 (W1-W8). Metrics:

• DECA Score: 0-100 index measuring recommendation frequency and prominence per brand per category per week
• Gini coefficient: Inequality measure applied to DECA Score distributions within each category
• AI Visibility Rate: (AI-recommended brands ÷ G2-listed brands) × 100
• Data point: One brand-category-week observation containing DECA Score, rank, and mention status. The 4,123 total reflects the sum of all brand observations across 10 categories over 8 weeks.

Market type classification: Monopoly (Gini ≥ 0.73), Oligopoly (Gini 0.65-0.73), Distributed (Gini < 0.65) G2 brand counts: Sourced from G2.com category pages, as of May 19, 2026. Limitations:

• Tracking covers ChatGPT only; other AI systems (Claude, Gemini, Perplexity) may show different patterns
• The GPT-5.5 transition happened mid-tracking, which introduces a variable that cannot be fully isolated
• 8 weeks provides directional trends but may not capture longer-term cyclical patterns
• DECA Score measures recommendation strength, not end-user behavior or conversion
• Prompt templates and collection parameters were held constant across all 8 weeks; however, model behavior can vary across sessions and user contexts
• Model labels (GPT-5.4, GPT-5.5) reflect the visible model identifier in ChatGPT at the time of each weekly collection, not independently verified model versions

External data sources cited:

• OpenAI official announcements: GPT-5.5 (April 23, 2026), GPT-5.5 Instant (May 5, 2026)
• TechCrunch: ChatGPT 900M weekly active users (February 2026)
• Profound: ChatGPT Branded Link Update analysis (May 7, 2026)
• Goodie: AI Search Traffic Report 2026
• Onely: AI product recommendation research

DecaGEO publishes weekly DECA Score rankings across 10 SaaS categories. For methodology details and category-level data, see the category structure framework and AI Visibility analysis.

---

​

Sources

• AI Visibility Has Two Filters: Why 91.6% of Software Brands Never Get Recommended — Background on the Two-Filter model for AI recommendation analysis
• GEO Strategy Depends on Category Structure: Why the Same Playbook Produces 2× Different Results — Category structure analysis framework
• GEO Opportunity Depends on Category Openness: 4 AI Recommendation Markets Are Still Open, 6 Are Already Closed — Market openness classification
• OpenAI: Introducing GPT-5.5 — Model release announcement
• OpenAI: GPT-5.5 Instant — Default model transition
• TechCrunch: ChatGPT reaches 900M weekly active users
• Profound: ChatGPT Branded Link Update
• Goodie: 2026 AI Search Traffic Report
• Onely: How to Get Your Product Discovered by ChatGPT