About
Why trust this database?
BetterAISearch exists because most “GEO tips” online are opinions dressed as facts. Every tactic in this database is scored against a four-factor evidence model — and we show the sources, methodology, and confidence score for every claim.
The problem
When AI search emerged, an entire content industry followed — blog posts claiming to know how to rank in ChatGPT or Perplexity, almost none of it backed by data. Marketers were left guessing: which tactics are actually confirmed? Which are vendor marketing? Which contradict each other?
BetterAISearch is a knowledge base that tracks GEO research as it publishes, scores it by source quality, and surfaces only what the evidence actually supports. If a finding has fewer than 3 independent sources, we say so. If a source contradicts a tactic, we show that too.
How we score research
Every tactic receives a confidence score based on four factors. The score determines the confidence label shown on each tactic page.
Source credibility
Tier 1 sources contribute the most to a confidence score. Each tier is weighted differently — platform-official documentation anchors the score, independent studies corroborate it.
Independence
We check whether sources have commercial interests that could bias the finding. Tier 4 sources are discounted proportionally. A tactic backed only by a vendor selling that service is flagged as low confidence regardless of study quality.
Freshness decay
GEO platforms change every 2–3 months — Google AIO, ChatGPT Search, and Perplexity each underwent major behavioural shifts multiple times in 2024–25. We calibrated the decay to match empirical data: 50% of AI-cited content is under 13 weeks old, making the effective research half-life roughly 3 months. A study from 18 months ago carries just 40% of its original weight. Platform-official sources (Tier 1) are exempt — their documentation defines how their own systems work.
Consensus
Multiple sources agreeing pushes a tactic toward Very High confidence. A single strong source, even Tier 1, is rated High at best. Contradicting evidence reduces the score proportionally — we show it rather than hide it.
Freshness decay schedule
Tier 1 platform sources (Google, OpenAI, Perplexity, Microsoft) are exempt from decay — their documentation defines how their own systems work. All other tiers decay aggressively because GEO platforms shift behaviour every 2–3 months. Calibrated against real citation data: 50% of content cited by AI engines is under 13 weeks old (Amsive, 2025).
Source tiers
Not all sources are equal. We classify every source into one of four tiers before scoring. The tier determines how much weight a finding contributes to the confidence score.
Direct statements from Google, Perplexity, OpenAI, or Microsoft. Published in official documentation, engineering blogs, or developer guidelines. These are the only sources that can definitively confirm what a platform's algorithm considers.
Examples: Google Search Central, OpenAI's developer docs, Perplexity's engineering blog, Microsoft Advertising guidelines.
Peer-reviewed studies, pre-prints on arXiv, or research from university labs with documented methodology. We require sample sizes, methodology description, and reproducibility. A study on 1,000 queries carries more weight than one on 50.
Examples: Princeton KDD GEO Study (2024), arXiv AI citation analysis, Stanford credibility research.
Controlled experiments run by practitioners with documented methodology, clear test vs. control setup, and verifiable results. Must be independent of vendors with commercial interests. Anecdotal case studies without controls are excluded.
Examples: Independent SEO A/B tests with documented sample sizes, Wix AI Search Lab experiments.
Reports from Ahrefs, Semrush, BrightEdge, and similar. Useful for scale and data access, but discounted for potential commercial bias and methodology opacity. Never used as sole confirmation of a tactic.
Examples: Semrush GEO Report, BrightEdge AI Search Report, Ahrefs citation analyses.
What we exclude
- ×Agency blog posts without documented methodology or data
- ×Anecdotal claims without a control group or before/after measurement
- ×Findings contradicted by platform-official documentation
- ×Research older than 24 months without recent corroboration
- ×Any claim where the source has a clear commercial interest in the outcome, uncorroborated by independent sources
Transparency
Every tactic page shows the full source list with direct links, the confidence score breakdown, and any contradicting evidence. We require a minimum of 3 independent sources before confirming a tactic — if the evidence isn't there yet, we say “insufficient evidence” rather than present weak findings as fact.
The database is updated as new research publishes. Tactics are re-scored when new evidence emerges — including evidence that contradicts existing findings.