AI Citation Trends: March 2026

As of March 2026, AI search engines disagree on the #1 brand recommendation in 50% of B2B software queries, and the disagreement is not shrinking. Across three weekly waves of identical queries sent to ChatGPT, Perplexity, Gemini, Grok, and Claude (300 engine-query pairs total), cross-engine consensus oscillated at 50%, 55%, 50%. The engines are not converging on stable answers. They are oscillating.

This is the inaugural monthly analysis from FogTrail's "State of AI Citations" research program: 20 queries, 5 B2B SaaS categories, 25 tracked brands, 3 waves collected between March 6 and March 15, 2026. Below is the full dataset, broken into key metrics, source type patterns, category deep dives, and the five trends that defined the month.

Key Metrics at a Glance

These 10 metrics summarize the state of AI citations across all engines for March 2026.

Source Type Distribution

Each engine pulls from different source categories, and those preferences are stable across waves. ChatGPT links to brand-owned websites at 4x the rate of Perplexity, Gemini, or Claude. Grok favors third-party review sites. Three engines (Perplexity, Gemini, Claude) route over 85% of their URLs to blogs, docs, and other non-branded content.

ChatGPT is the only engine that cites Wikipedia with any regularity (5.3% of its URLs). ChatGPT's brand-owned URL share is also declining wave over wave (24.2%, 19.0%, 18.4%), while Grok's is climbing (1.9%, 6.0%, 8.5%). If both trends hold, they will converge near 10-15% brand-owned URLs by mid-year, while Perplexity, Gemini, and Claude remain below 5%.

For brands, this means that optimizing your own website matters most for ChatGPT and increasingly for Grok. On the other three engines, third-party content, blog posts, and documentation drive citations.

The Five Findings That Defined March

1. AI Engines Oscillate Rather Than Converge

This is the single most important finding of March. The wave-by-wave consensus data kills the assumption that AI engines are gradually aligning on stable answers.

The Wave 2 "convergence" signal (55% consensus, 63% overlap floor) was a one-wave artifact. Wave 3 reverted to Wave 1 values exactly. Any brand tracking AI visibility from a single snapshot is measuring noise, not signal.

2. PostHog Is the Only Brand With 3 Consecutive Waves of Citation Growth

PostHog citations grew from 2 (W1) to 3 (W2) to 5 (W3), making it the most-cited startup in the dataset and matching Salesforce's citation count in Wave 3. It is now cited by all 5 engines, the only startup to achieve this.

No other brand shows three consecutive increases. PostHog's open-source model and strong documentation appear to be generating durable AI search signals. More on this in our analysis of startups outranking category leaders in AI search.

3. ChatGPT Is Becoming the Outlier Engine

ChatGPT has the lowest pairwise overlap with 3 of the other 4 engines. Its overlap with Gemini (58% in Wave 3) is the lowest pair in the entire three-wave dataset. ChatGPT is also the most volatile engine by citation count (23, 12, 14 across waves) and the only engine to use Wikipedia heavily (10.4% of URLs in W3).

ChatGPT is selectively aligning with Grok while diverging from the Perplexity-Gemini-Claude cluster. The highest-traffic AI engine is also the most unpredictable.

4. Category Consensus Is Diverging at Dramatically Different Rates

Analytics consensus improved steadily from 1/4 to 2/4 to 3/4 queries with strong agreement. Project Management stayed stuck at 0/4 for two consecutive waves. CRM, previously one of the most stable categories, cracked from 3/4 to 1/4 in a single wave.

The gap between the most consolidated category (Analytics at 75% consensus) and the most fragmented (PM at 0%) is widening. For PM brands, the window of opportunity remains wide open. For analytics challengers, it is closing fast. Full breakdown in our category consensus divergence analysis.

5. Incumbent Advantage in "Alternative to X" Queries Stabilized at 87%

When users ask AI for alternatives to a product, the incumbent still gets position #1 in 87% of engine responses. This rate dropped from 93% in Wave 1 to 87% in Wave 2 and held at 87% in Wave 3.

The erosion was real (93% to 87%), but it has plateaued. The crack is exclusively in CRM, where HubSpot displaces Salesforce in 2 of 5 engines. Mailchimp and Google Analytics remain untouchable. More context in our post on alternative queries favoring incumbents.

Category Deep Dives

CRM

HubSpot and Salesforce are locked in a genuine position war. Both achieve 5-engine coverage and near-identical citation counts, but HubSpot wins more #1 positions for startup-focused queries while Salesforce dominates on Claude (3/4 CRM queries with Salesforce at #1 across all waves). CRM consensus cracked sharply in W3, dropping from 3/4 to 1/4, driven by Perplexity and Grok rotating their CRM preferences. Close rotates which engines surface it each wave. Attio's regression from 1 mention to 0 for two consecutive waves confirms it is genuinely invisible.

Project Management

PM is the most contested category in the dataset. No PM brand has achieved 3/5 or better on any query in any wave. Asana, Monday.com, ClickUp, and Linear rotate #1 positions constantly. ClickUp has strong mentions (15 per wave average) but zero citations across all three waves, making it the most "mentioned but never linked" brand. Height's 300-pair invisibility is the most extreme case in the entire study. Perplexity and Claude consistently return zero tracked PM brands for "engineering PM" queries, recommending Jira and Microsoft Project instead.

Email Marketing

Mailchimp dominates overall, but Beehiiv is winning the newsletter sub-niche. Beehiiv's consensus grew from 2/5 to 3/5 across three waves as Gemini flipped from Mailchimp to Beehiiv for newsletter queries. ActiveCampaign's complete disappearance from ChatGPT in Wave 3 (from 2 citations to 0) is the most dramatic single-engine brand drop in the dataset. It still appears on 4 of 5 engines despite the ChatGPT loss, ruling out a universal signal change.

Analytics

Analytics is the only category where all 5 brands appear on all 5 engines in every wave. Consensus improved steadily (1/4, 2/4, 3/4), with Amplitude emerging as the consensus leader for product analytics (5/5 unanimous on comparison query in W3). Google Analytics has 16 mentions per wave but zero citations across all three waves, likely because engines describe it without needing to link to it. PostHog's citation growth (2, 3, 5) is the standout startup success story of the month.

Dev Tools

Vercel dominates position #1 but its rate is slowly declining (100%, 94%, 88%). Netlify broke its 0-for-28 streak in Wave 3 when ChatGPT placed it first for "deploying web apps," though this happened only on ChatGPT, the most volatile engine. Railway has strong mentions (9 per wave) but zero citations across all waves. Whether Netlify's breakthrough persists will be a key signal in April.

Five Trends to Watch

1. Engines Form Shifting Alliances, Not a Unified Consensus

Pairwise overlap data reveals that engines cluster differently each wave. In Wave 1, Perplexity-Gemini had the highest agreement (71%). In Wave 2, Grok-Gemini surged to 79%. In Wave 3, Grok-Claude became the highest pair (75%). There is no stable "consensus cluster." The most durable relationship: Grok-Claude convergence has risen steadily (62%, 72%, 75%). The most volatile: ChatGPT-Gemini has swung widely (62%, 67%, 58%).

2. Citation Counts Are Engine Fingerprints, Not Quality Signals

Each engine has a remarkably stable URL generation pattern, regardless of what it is recommending.

Grok produces 3.5x the URLs of ChatGPT, but this does not translate to more brand citations. Grok's brand citation efficiency is the lowest of all engines. The volume is noise, not signal.

3. Brand-Owned URL Share Is Diverging by Engine

ChatGPT links to brand websites at a declining rate (24.2%, 19.0%, 18.4%), while Grok is climbing (1.9%, 6.0%, 8.5%). All other engines are flat or declining. If these trends continue, Grok and ChatGPT will converge near 10-15% brand-owned URLs, while the other three engines remain below 5%. Optimizing your own website content matters most for ChatGPT and increasingly for Grok.

4. The Startup Ceiling Is Real but Has Exceptions

No engine exceeds a 15% startup-at-#1 rate in Wave 3 (down from ChatGPT's 25% in Waves 1-2). Enterprise brands hold position #1 in 45-60% of queries depending on the engine. The structural advantage of being large persists in AI search. However, individual startups can break through: PostHog (5 citations, 5 engines), Beehiiv (3/5 newsletter consensus), and Linear (consistent PM presence) demonstrate that niche positioning can overcome the size disadvantage.

5. Claude Is the Most Deterministic Engine

Claude produced exactly 6 brand citations in all three waves. Its brand coverage barely shifted. Claude recommended Salesforce first for "CRM for startups" in all three waves, dissenting against a 4-engine HubSpot consensus each time. For brands, Claude is the most predictable engine: if you are invisible on Claude today, you will likely still be invisible next week. Its volatility score (8%) is the lowest of any engine, compared to ChatGPT's 39%.

What You Can Do About It

• Monitor multiple engines, not just one. The engines disagree on the top recommendation 50% of the time. Single-engine tracking gives you half the picture at best.
• Track over multiple waves. ChatGPT's citation count swung by 48% between consecutive weeks. A single snapshot is noise. Three or more observations establish signal.
• Prioritize niche positioning over broad authority. PostHog, Beehiiv, and Linear all broke through by owning a specific sub-category (open-source analytics, newsletters, dev-focused PM) rather than competing head-on with incumbents.
• Optimize brand-owned content for ChatGPT and Grok specifically. These two engines link to brand websites at meaningfully higher rates than the other three, and Grok's rate is climbing.
• Check "alternative to [competitor]" queries. Incumbents hold position #1 in 87% of these responses. If you are the challenger, this is the hardest query type to win, but CRM shows it can crack.

Methodology

We ran 20 queries across 5 AI search engines (ChatGPT, Perplexity, Gemini, Grok, and Claude) once per week for three consecutive weeks (March 6, March 10, March 15), producing 300 engine-query pairs. Each query was sent as a real-time API call with no temperature overrides, simulating how actual users interact with these platforms. We tracked 25 B2B SaaS brands across 5 categories (CRM, Project Management, Email Marketing, Analytics, Dev Tools), classified as enterprise (6), midmarket (8), or startup (11).

A citation is defined as a direct hyperlink to a tracked brand's owned domain within an engine response. A mention is the brand named in response text, with or without a link. Consensus means 4+ of 5 engines agree on the same #1 brand for a query. Pairwise overlap uses the Jaccard index of brand mentions shared between two engines across all 20 queries. All engines use temperature-based sampling. Patterns stable across 3+ queries or multiple waves are considered signal. Single-mention appearances are considered noise.

FAQ

How often do AI engines agree on the best B2B software recommendation?

As of March 2026, AI engines reach strong consensus (4+ of 5 agreeing) on only 50% of B2B software queries. Unanimous agreement (5/5) occurs in roughly 27% of queries. The consensus rate oscillated between 50-55% across three weekly waves with no upward trend.

Which AI engine is the most volatile for brand citations?

ChatGPT is the most volatile engine by a wide margin. Its brand citation count swung from 23 to 12 to 14 across three identical weekly runs, a volatility score of 39%. Claude is the most stable at 8% volatility, producing exactly 6 brand citations in every wave.

Can startups compete with enterprise brands in AI search results?

Startups are cited at a 58% rate compared to 100% for enterprise brands, but individual startups can break through with niche positioning. PostHog grew from 2 to 5 citations across three waves by focusing on open-source analytics. Beehiiv achieved majority consensus for newsletter queries. The structural advantage of size persists, but it is not absolute.

Do AI engines cite Reddit as a source?

There is a clear two-tier divide. Grok (30% of queries) and ChatGPT (25%) regularly cite Reddit URLs. Gemini (5%), Perplexity (2%), and Claude (0%) almost never do. This pattern held stable across all three waves.

What happens when users ask AI for alternatives to a market leader?

The incumbent retains position #1 in 87% of engine responses. This rate dropped from 93% in the first wave but stabilized at 87% for two consecutive waves. The erosion is exclusively in CRM. Mailchimp and Google Analytics maintain 100% incumbent advantage across all engines and all waves.

Related Resources

• We Asked 5 AI Engines the Same 20 Questions. They Disagreed on the #1 Answer 50% of the Time.
• We Ran the Same 20 Queries 3 Times Across 5 AI Engines. Here's How Much the Results Changed Each Week.
• ChatGPT Agrees With Other AI Engines Less Than 60% of the Time on Brand Recommendations
• Analytics Consensus Surged From 25% to 75% While Project Management Is Stuck at 0%
• 3 Startups That Outrank Their Category Leaders in AI Search