Most brands are losing AI citations and don't even know it
Only 14% of marketers track whether their brand appears in AI answers — yet AI Overviews serve different content 70% of the time for the same query, and the same brand list almost never repeats across runs. Your brand could be disappearing from AI recommendations right now and you would have no idea.
Almost no one is watching
A GoodFirms survey of 100+ marketing practitioners across 20+ countries (Jan–Feb 2026) found that just 14% of marketers actively monitor AI citation visibility. That number sits against a backdrop where 89% of brands already appear in AI search results and 43% of the same practitioners name AI search optimisation as a core 2026 strategy. The gap between ambition and measurement could hardly be wider. Tracking adoption is also sharply size-dependent: 73% of marketers at firms with 1,000+ employees monitor AI visibility, versus 39% at firms with ten or fewer — according to Fractl's AI Search study of 150 marketers.
AI answers are reshuffling constantly
Ahrefs studied roughly 43,000 AI Overview keywords and found that the content changes ~70% of the time for the same query — and 45.5% of individual citations change when an AI Overview regenerates. That means a brand present in an answer today may be gone tomorrow through no action of its own, and vice versa. At the individual brand-list level, the numbers are even more striking. SparkToro and Gumshoe ran 12 prompts across roughly 2,961 total runs on ChatGPT, Claude and Google AI (600 volunteers, Nov–Dec 2025): the same brand list appeared in fewer than 1-in-100 runs, and the same list in the same order appeared in fewer than 1-in-1,000.
Which AI you ask changes which brands get cited
It is not just that answers vary over time — the sources cited vary radically by platform. Ahrefs found only 13.7% of citations overlap between Google AI Overviews and Google AI Mode. A separate Profound analysis of 6.8 million citations found that only about 11% of domains cited by ChatGPT also appear in Perplexity results. A brand can be prominent in one AI ecosystem and virtually absent in another, with no single dashboard capturing both. The table below illustrates the scale of divergence.
| Comparison | Citation overlap | Source |
|---|---|---|
| Google AI Overviews vs Google AI Mode | 13.7% | Ahrefs |
| ChatGPT vs Perplexity (domain level) | ~11% | Profound (6.8M citations) |
| Same AI Overview query regenerated | 54.5% retained | Ahrefs (~43K keywords) |
| Identical brand list across ~2,961 runs | <1% | SparkToro + Gumshoe |
Large brands monitor. Small brands fly blind.
The Fractl study (1,008 consumers + 150 marketers) shows monitoring adoption breaks sharply on company size. Enterprise teams have the resources to track AI visibility; smaller brands — the ones arguably most exposed to competitor displacement from a single AI answer — largely do not. Meanwhile, 27% of marketers across the full sample report their brand has already been inaccurately described in an AI response. Without monitoring, that inaccuracy goes uncorrected and compounds across every query that surfaces it.
| Company size | Monitor AI citation visibility |
|---|---|
| 1,000+ employees | 73% |
| 10 or fewer employees | 39% |
| All sizes (GoodFirms) | 14% |
What the research does not tell you
A few important caveats before acting on any of this — including Baseline's own platform data.
Rank tracking in AI is probably meaningless. SparkToro's conclusion from their research is pointed: position-level tracking — "my brand moved from third to second" — is "statistically meaningless" given how rarely lists repeat. The more stable signal is share-of-presence: how often your brand appears at all across many independent runs, not where it ranks in any single one.
Tools use incompatible methods. Prompt-tracking, screenshot sampling, and direct API calls each measure something slightly different. Numbers from different tools are not directly comparable. Ahrefs itself flags that part of its measured citation "change" may reflect improvements in how it parses AI Overview content, not genuine volatility in the answers.
Vendor incentive. The majority of research quantifying how poorly brands track AI citations — including the GoodFirms and Fractl studies most cited here — comes from companies that sell monitoring or adjacent services. Baseline sells monitoring too. That does not make the numbers wrong, but it is worth factoring in when deciding how urgently to act.
Measure presence, not position
Given the volatility, the right unit to track is share-of-presence across many runs on many platforms — not where you ranked in one answer on one day. That means running the same prompts repeatedly, across ChatGPT, Perplexity, Google AI and Claude, and tracking how consistently your brand appears in the pool of recommended names. A single snapshot is noise. A trend across dozens of runs is signal. The 14% of marketers who are already doing this are building a lead that will compound as AI search share grows.
Sources: GoodFirms — SEO statistics, AI search rankings and zero-click trends (100+ practitioners, Jan–Feb 2026, published Apr 7 2026). Ahrefs — How often do AI Overviews change? (~43,000 keywords). Ahrefs — AI Overviews vs AI Mode. SparkToro + Gumshoe (Rand Fishkin / Patrick O'Donnell) — AIs are highly inconsistent when recommending brands or products (600 volunteers, ~2,961 runs, Nov–Dec 2025). Fractl — AI Search study (1,008 consumers + 150 marketers; author Aditya Sachdeva). Profound — 6.8M citation analysis (cited via Ahrefs AI Mode post).