Tremor measures how unusually each AI model changed its answers from one week to the next. For the same set of prompts, we compare the sources cited, the brands mentioned and the sentiment expressed, score each signal against that model's own recent baseline, and blend them into a single 0-10 magnitude per model.

What does a high magnitude mean?

AI answers naturally vary a little every week, so Tremor scores each model against its own typical movement. A magnitude around 2 is a normal week. A high magnitude means the model's answers shifted far more than usual, with different sources, brands appearing or disappearing, or changing sentiment. Near 0 means its answers were unusually still.

How often is Tremor updated?

It updates every Tuesday. Most weekly AI answers are refreshed on Monday, so by Tuesday the new week's data has landed and we can compare it to the previous week.

Which signals go into the magnitude?

Three signals: citation churn (how much the cited sources changed), mention churn (brands appearing, disappearing or moving position) and sentiment variation (how the tone toward brands shifted). Each signal is scored against that model's trailing eight-week baseline, and citations carry the most weight.

Where does the data come from?

It is built on the LLM Pulse dataset of millions of real AI answers across ChatGPT, Perplexity, Gemini, Google AI Mode and AI Overviews. The same data powers brand-level tracking inside LLM Pulse.

Tremor by LLM Pulse

The AI Search Volatility Tracker

A weekly seismograph for AI search. Tremor measures how unusually each AI model changed its answers week-over-week, comparing citations, brand mentions and sentiment to each model's own baseline, as one magnitude from 0 (calm) to 10 (major). Updated every Tuesday.

Week of Jul 13, 2026

This week across all AI search

2.4 / 10

Minor ▲ +0.7 vs last week

0 · Calm 5 10 · Major

Based on 201,002 repeated prompts

This week by AI model

Each model gets its own magnitude, scored against its own recent baseline. Higher means its answers, sources and brand mentions shifted more than usual since last week.

AI search volatility: weekly history

Tremor magnitude (0-10) per AI model, week-over-week. Updated every Tuesday by LLM Pulse.

Reading the scale

Calm

0–2

Minor

2–4

Moderate

4–6

Strong

6–8

Major

AI-referred traffic trends

How AI-referred visits are trending across connected sites, as an anonymized index per engine (start of window = 100). Relative change only, no traffic volumes.

Anonymized index across connected analytics accounts (k-anonymized; volumes never exposed). Updated weekly by LLM Pulse.

Privacy by design: each site is normalized to its own baseline, then aggregated as a median across many sites. No individual site or visit count is ever shown.

What Tremor measures

Mozcast and Algoroo gave SEOs a weekly read on how much Google's algorithm moved. Tremor does the same for AI search, but instead of one engine, it tracks how much each AI model changes its answers from week to week.

Every week we take the prompts that ran in both the current and the previous week, and for each AI model we measure how much the cited sources, brand mentions and sentiment shifted. Because AI answers always move a little, each signal is scored against that model's own recent baseline, then blended into a single seismic magnitude from 0 (unusually calm) to 10 (major upheaval). A normal week reads around 2.

How we measure it

Citation churn: for the same prompts, we compare the set of domains each model cited this week vs last week (Jaccard distance). This is the strongest signal, the AI-search equivalent of "the results changed".
Mention churn: how often brands appeared, disappeared or moved position in answers week-over-week.
Sentiment variation: how much the tone toward brands shifted week-over-week.
Baseline-relative scoring: each signal is compared to that model's own trailing eight-week baseline (robust z-score), so the magnitude measures unusual movement, not the natural week-to-week noise of AI answers. A typical week reads around 2; higher means the model moved beyond its own usual volatility. Signals are blended (citations weighted most) into the headline magnitude, and the `all` reading averages the per-model magnitudes.
Updated every Tuesday, comparing the latest complete week to the one before it. Built on the LLM Pulse dataset across ChatGPT, Perplexity, Gemini, Google AI Mode and AI Overviews.

Why it matters

A spike in Tremor means an AI model just reshuffled who it cites and recommends, a leading indicator that your brand's AI visibility may have shifted, for better or worse.

Tracking volatility by model tells you where to look first: a quiet week on ChatGPT but a storm on Google AI Mode means your AI search risk is concentrated in one place.

Tremor: Frequently Asked Questions

: Tremor measures how unusually each AI model changed its answers from one week to the next. For the same set of prompts, we compare the sources cited, the brands mentioned and the sentiment expressed, score each signal against that model's own recent baseline, and blend them into a single 0-10 magnitude per model.
: AI answers naturally vary a little every week, so Tremor scores each model against its own typical movement. A magnitude around 2 is a normal week. A high magnitude means the model's answers shifted far more than usual, with different sources, brands appearing or disappearing, or changing sentiment. Near 0 means its answers were unusually still.
: It updates every Tuesday. Most weekly AI answers are refreshed on Monday, so by Tuesday the new week's data has landed and we can compare it to the previous week.
: Three signals: citation churn (how much the cited sources changed), mention churn (brands appearing, disappearing or moving position) and sentiment variation (how the tone toward brands shifted). Each signal is scored against that model's trailing eight-week baseline, and citations carry the most weight.
: It is built on the LLM Pulse dataset of millions of real AI answers across ChatGPT, Perplexity, Gemini, Google AI Mode and AI Overviews. The same data powers brand-level tracking inside LLM Pulse.

Track your own brand

See how AI search shifts for your brand, every week

Tremor is built on the LLM Pulse dataset. Track citations, mentions, sentiment and share of voice for your own brand across every AI search engine.

Citations, mentions and sentiment for your brand and competitors
Every major AI model, including ChatGPT, Perplexity, Gemini, Google AI
Weekly tracking, share of voice and alerting

Start Free Trial Browse Data Studies

The AI Search Volatility Tracker

This week by AI model

ChatGPT

Perplexity

Gemini

Google AI Mode

Google AI Overviews