See what's shifting in AI search.

SeerSonar tracks a fixed set of buyer questions in ChatGPT across 5 core industries and measures how much its answers, sources, and formatting move over time, so you know when AI search is changing and who's gaining or losing visibility as it does. 50,707 responses and counting.

Talk to our GEO team → How Seer approaches GEO →
Current readingiOne 0–100 score for how much ChatGPT's answers to the same set of questions moved in the last full week, measured against a typical week from the first half of 2026. Low = a calm week; Extreme = a lot shifted.
 
 

Loading the latest reading…

 
LowExtreme
 
Date range

Volatility over timeiHow much ChatGPT's answers change between two back-to-back runs of the same question. Combined is one blended score; Breakdown splits it into the three signals it's built from.In Breakdown, all three lines share one scale — 50 is a typical week — so you can read them against each other.

Volatility tracks how much ChatGPT's answers move between consecutive runs of the same prompt. The score blends citation, structural, and content signals, with citation weighted heaviest. Switch to Breakdown to see each signal on its own.

 
 
Dashed lines mark OpenAI/ChatGPT model releases. This is shown as context for what may have moved, not proof of cause. Grey bands mark thin weeks, which is defined by a gap in collection data, not necessarily a quiet week.

Response elementsiHow often each formatting element (bold, tables, lists, prices, and so on) shows up in answers — the “SERP features” of AI answers.Different from Content volatility above, which measures wording change, not formatting.

The "SERP features" of AI answers show how often each element — like lists, tables, and text formatting — shows up and how those trends change over time.

 
 
 
 

Citation rate over timeiShare of answers that cite at least one external source.Different from Citation volatility — this is whether it cites, not how much the sources change.

Share of ChatGPT answers that cite at least one external source. Paired with the cited-domains view below so you can move from "how often it cites" straight to "who it cites."

 Answers citing a source
 

Cited domainsiWhich sites ChatGPT cites most, and how each site's share of citations moves over time.

Which sources ChatGPT cites most. Pick an industry to scope the list, then click any domain to see its citation share over time — where the real gains and losses show up.

 
 
 
 
 

The Seer Newsletter  |  Bimonthly

We track the shifts, you stay ahead.

What we're learning from SeerSonar and the rest of our research, plus our unfiltered take on what it means for your visibility. Free, twice a month, no fluff.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

What we measure

Three signals, each measured between consecutive runs of the same prompt, blended into one reading — weighted by how much real movement they carry (citation ~50%, structural ~35%, content ~15%), not treated as equal.

Primary
Citation volatility

How much the set of cited sources changes run to run. The lowest-noise, most business-relevant signal — it's who wins and loses visibility — so it carries the most weight.

Event signal
Structural volatility

Whether formatting elements (headings, tables, lists) appear or disappear. Usually near-flat, so when it jumps it's a reliable marker of a genuine change in how ChatGPT answers.

Light weight
Content volatility

How reworded the answer is. ChatGPT rewrites ~90% of every answer regardless (it barely moves), so it's folded in at a low weight (~15%), normalized so its flatness can't dominate — it only nudges the reading when rewording genuinely spikes.

How it works

01
What we track

A fixed set of 625 buyer-style prompts across 5 industries (Technology, Healthcare & Life Sciences, Finance & Insurance, Real Estate & Hospitality, and Industrial & Manufacturing), re-run in ChatGPT roughly every three days since January. Keeping the set fixed is what lets us compare today to last week and last month.

02
What we measure

For each prompt we compare each run to the one before it — the sources cited, the formatting, and the wording — and roll the movement into a single reading, weighted toward the signals that carry real change.

03
How we read it

The reading is measured against a fixed baseline — the first half of 2026 — so a typical week sits mid-scale, spikes are meaningful rather than arbitrary, and a week's number never changes once it's published. Every chart shows the movement the reading is built from — nothing is a black box.

04
Why it matters

Visibility inside AI answers moves far more than a search ranking does. SeerSonar is the instrument for "how much did AI search shift this week, and who moved?" — and the drill-downs are where the discoveries live.

SeerSonar · Seer Interactive. A fixed set of 625 prompts is run repeatedly in ChatGPT; each run is compared to the previous one across citations, structure, and content, then measured against a fixed baseline from the first half of 2026 into a single volatility reading. Figures on this page are computed from live study data (Jan–Sep 2026).

Read more about what we're studying, measuring & tracking in AI Search

Ai

Re-imagining AI Visibility KPIs in a Query Fan Out world

Wil, take me right to the KPI for long term sustainable GEO visibility!!
Ai / Geo

AI Visibility Is Lying to You: There’s No One-Size-Fits-All AI Search KPI
Ai

AI Citations Aren't Your Top KPI: ChatGPT-5.6 Luna Runs More Targeted Searches
Ai Study

Study: AI Declined to Answer Prompts About Our Brand 31% of the Time (But Got Our Info Mostly Right When It Answered)