Why AI Brand Scores Are Volatile (And What That Means for Marketers)

Last week, we ran a brand check for Allbirds on Perplexity.

First run: 17/100 — negative sentiment.
One hour later: 100/100 — positive sentiment.

Same brand. Same day. Completely opposite results.

This isn't a bug. It's a feature of how AI search engines work — and it has serious implications for anyone tracking brand visibility in AI.

Why AI Search Results Aren't Deterministic

Traditional search engines (Google, Bing) serve cached, indexed results. The same query returns the same ranking today and tomorrow — until an algorithm update changes it.

AI search engines are different in three key ways:

1. Retrieval-augmented generation (RAG)

Engines like Perplexity actively fetch and synthesize live web content at query time. If a negative news story about your brand was recently published, it may be included in the synthesis — or not, depending on retrieval randomness.

2. Temperature and sampling

Large language models generate responses probabilistically. Even with identical prompts and retrieved documents, the response can vary between runs.

3. Real-time index updates

Perplexity's index refreshes continuously. A brand mentioned in a viral Reddit thread at 9am may appear prominently at 10am and fade by noon as newer content crowds it out.

The Practical Impact

We've now run 200+ brand checks across ChatGPT, Gemini, Perplexity, and Claude. Here's what we've observed:

Engine	Score Stability	Why
ChatGPT	High	Mostly trained on static corpus; less real-time retrieval
Gemini	High	Google Knowledge Graph integration provides consistent entity data
Perplexity	Low	Heavy RAG with live index; most volatile engine
Claude	Medium	Trained data + some retrieval; more consistent than Perplexity

Perplexity is the most volatile engine by far. If your brand has inconsistent coverage across web sources, Perplexity will surface that inconsistency — and its score can swing dramatically between checks.

What This Means for Your Brand

Score volatility = narrative fragility

If your brand's AI visibility score is stable across engines and over time, it means your brand narrative is deeply embedded and consistent. High-volume, authoritative coverage from multiple independent sources is hard to displace.

If your score is volatile — especially on Perplexity — it typically means:

• Your web coverage is thin or concentrated in a few sources
• Your narrative is inconsistent across sources
• Recent events (a bad review, a news story) are disproportionately influencing AI responses

The monitoring implication

A single brand check is a snapshot. It tells you what AI said about you at one point in time.

Monitoring — checking weekly or monthly — tells you:

• Whether your narrative is improving or degrading
• If a specific engine is shifting (often the first signal of a PR or reputational issue)
• The impact of your GEO efforts over time

How to Stabilize Your AI Brand Score

The antidote to volatility is the same regardless of engine: volume and authority.

1. Wikipedia presence — the most consistently cited source across all AI engines
2. Press coverage — articles in major publications get weighted heavily
3. Community discussions — Reddit, Hacker News, and Indie Hackers create persistent narrative signals
4. Case studies and data — third-party validation of your claims
5. Consistent messaging — avoid contradicting your own narrative across channels

The brands that score consistently high (Basecamp 100/100, HubSpot 100/100) have one thing in common: they've been saying the same things about themselves for years, and the internet has been agreeing with them for years.

Check Your Score

You can run a free brand check at geo.atlas1m.com — no signup needed. It runs your brand name against ChatGPT, Gemini, Perplexity, and Claude simultaneously and returns scores, sentiment, and the full AI response text.

Run it twice on the same day and compare. If the scores are stable, your narrative is solid. If they swing — you have a GEO problem worth fixing.