Neil Patel published a piece last month titled [GEO Best Practices: Prompt Volume Shouldn't Drive Strategy](https://neilpatel.com/blog/geo-best-practices-prompt-volume-shoudnt-drive-strategy/). He's right. Prompt volume — the number of times a query is asked in ChatGPT, Perplexity, or Google AI Mode — is a lagging, low-signal metric that tells you almost nothing about whether optimising for that query will pay off. His fix is to prioritise "buyer intent". We agree with the diagnosis; we'd sharpen the prescription. The metric that beats prompt volume isn't intent — it's **citation probability**.
What citation probability actually is
Citation probability is a per-query estimate of how likely an AI answer engine is to cite your page if you ship one on that topic. It's a function of four things:
- **Incumbent quality** — how strong the currently-cited page is on the [GEO Standard signals](/standard). A citation held by a thin, undated page is a citation up for grabs.
- **Structural fit** — whether the query rewards answer-first structure, listicle format, or long-form explanation. Different engines reward different shapes.
- **Freshness runway** — how quickly the topic decays. AI models revert to freshest-available content aggressively; slow-moving topics have longer citation half-lives.
- **Your delta** — the honest gap between what you can ship and what the incumbent has. If you can only match, don't bother.
A query with 40,000 monthly prompts and a Wikipedia citation has a citation probability near zero. A query with 800 monthly prompts and a 3-year-old thin blog post as the incumbent has a citation probability near one.
The 4:1 result
We ran the head-to-head across 22 sites for 90 days ([full methodology and data](/stats/citation-probability-beats-prompt-volume)):
- **Strategy A — Prompt volume.** Target the top 20 highest-prompt-volume queries in each site's category. 1,140 sign-ups across cohort.
- **Strategy B — Citation probability.** Target the top 20 queries where the incumbent citation scored <70/100 on the GEO Standard and the query had commercial intent. 4,610 sign-ups across cohort.
Ratio: **4.04:1** in favour of citation probability.
Neil's cohort of 22 companies saw search leads go from 3.1% to 7.4% year over year. That's real. Our claim is that most of the additional 4.3 percentage points came from a small number of citation-probability wins, not from broad prompt-volume alignment.
Why prompt volume misleads
High-volume queries look like opportunity. They're not — they're incumbent-dominated. The top 3 citations for any 20,000+ monthly-prompt query on our sample were owned by Wikipedia, an official brand page, or a 100k+ backlink editorial page 87% of the time. Displacing any of them is a 12–18 month campaign.
Medium and long-tail queries look like scraps. They're not — they're where the incumbent is a mediocre 2023 blog post and a well-structured page can flip the citation in 4 weeks.
Where Neil's framing works
Buyer intent is a real signal — but it's a filter, not a prioritisation. Use it to eliminate queries whose intent doesn't match your revenue (informational-only, unrelated categories). Then within the remaining set, rank by citation probability, not by prompt volume.
And Neil's 7-layer taxonomy is genuinely useful for stakeholder framing — enough that we built a full [side-by-side comparison](/vs/neil-patel-7-layer-stack) explaining when to use each framework. Different tools; both work.
Run it yourself in one afternoon
1. Pull your top 50 target queries by commercial intent. 2. Run each in ChatGPT, Perplexity, and Google AI Mode. Log the currently-cited URL. 3. Score each cited URL on [/check](/check). 4. Filter to queries where the incumbent scored <70/100. 5. That is your priority list. Ship a better page for each — answer-first, JSON-LD, freshness stamp, original data if you have it. 6. Re-measure at 4 weeks. Expect a 15–35% citation-share lift on your target queries; expect prompt-volume-ranked queries to be flat.
Where to go from here
- Read the full 4:1 study: [Citation probability beats prompt volume](/stats/citation-probability-beats-prompt-volume).
- Compare the two frameworks head-to-head: [Grow vs Neil Patel's 7-Layer Stack](/vs/neil-patel-7-layer-stack).
- Score any URL against the six GEO Standard signals: [/check](/check).