Lead with the conclusion
A concise lead answer recurs across three reviewed studies.
Open research · Open source · v2.0.1
Understand what makes content retrievable and citable.
In brief. This is a research-led guide to AI search citation readiness. AI search citation readiness is defined as a page's technical, retrieval, reuse, and provenance fitness for an engine-generated answer. aiseo-audit connects peer-reviewed research to a deterministic page audit without pretending its score predicts a citation.
npx aiseo-audit https://yoursite.comCurrent evidence
The evidence supports a short list of useful actions.
According to C-SEO Bench (NeurIPS 2025), source position mattered more than any fixed rewriting intervention it tested [1]. According to SAGEO Arena, structural-field and body changes affected different pipeline stages in its tests [2]. Together, the results support a staged audit instead of one universal formatting recipe.
Separate retrieval from citation fitness
Structural fields help a source enter context. Body evidence helps after retrieval.
Do not reward keyword repetition
Added keyword density hurt retrieval in tested benchmarks. The audit penalizes repetition but never rewards it.
Formatting counts are diagnostics
Lists, tables, and section counts support a review. Controlled tests do not justify points for their mere presence.
Engine behavior remains unstable
The same query produced different sources and citations under stable-looking settings.
Product pages need their own profile
Price, specifications, and comparisons affect product pages differently from informational pages.
Citation pipeline
One score, four different questions.
A strong answer still fails when the page never enters context. A page audit refers to a review of those visible stages. The pipeline keeps them separate instead of blending every signal into an unexplained number. The order matters because each later stage depends on the stage before it. This means teams fix access and retrieval gaps before polishing the answer. The reason for the split is measurement, which is why the score reports four stages.
- 01
Technical eligibility
Technical eligibility refers to page access, useful text extraction, and crawler access.
- 02
Retrieval alignment
Retrieval alignment refers to structural fields, terms, and entities that match the target query.
- 03
Citation fitness
Citation fitness means that retrieved content gives a direct, grounded, query-relevant answer.
- 04
Provenance
Provenance is defined as visible authorship, organization identity, and attribution.
Paper reviews
Primary sources, read paper by paper.
Each review records the experiment, metric, limitations, and the exact tool changes supported by the evidence. It also records what the evidence does not support.
NeurIPS 2025 Datasets & Benchmarks
C-SEO Bench: Does Conversational SEO Work?
Most fixed rewriting tactics did not transfer. Source position in context mattered more than any rewrite the study tested.ICLR 2026
What Generative Search Engines Like (AutoGEO)
Engine-specific, query-aware rules beat one universal checklist. The study's visibility metric limits transfer beyond its tested setting.KDD 2026
SAGEO Arena
Aligned structural fields improved retrieval. Body-only rewrites hurt retrieval in tested cases even when they helped generation.SIGIR 2026
What Gets Cited: Competitive GEO in AI Answer Engines
A controlled study ran 252,000 trials. Topic match and product price acted as gates, while other signals had smaller effects.Apply the research
Audit a real page, then inspect the evidence.
The command-line tool runs locally, needs no external AI key, and reports the evidence tier and paper citations behind each factor. Treat the result as a repeatable readiness audit. Do not treat it as citation probability.
$ npx aiseo-audit https://example.com$ aiseo-audit https://example.com --query "your target query"Methodology
How evidence enters the score
Factors are checked against primary research, assigned a tier, mapped to a pipeline stage, and blocked from release when their evidence record is missing.
Read the methodology →Project history
How the tool changed its mind
Version 2.0 removed unsupported score points, corrected freshness logic, and stopped presenting audit weights as additive citation gains.
Read the history →Limitations
What the audit cannot tell you
It does not render client-side content, inspect pixels, measure private engine indexes, or promise that a deployed model will cite a page.
Read the limitations →Key takeaways
What the current evidence supports.
- The pipeline is a sequence of technical eligibility, retrieval alignment, citation fitness, and provenance.
- The query is the audience question used to judge page alignment.
- The audit is a repeatable review of observable page and site signals.
- The factor is traceable to an evidence tier and source.
- The score is a readiness measure, not a citation forecast.
Bottom line: improve the earliest weak stage, keep the query set stable, and measure the page again.
References
- C-SEO Bench: Does Conversational SEO Work? : NeurIPS 2025 Datasets & Benchmarks.
- SAGEO Arena : KDD 2026.
- What Gets Cited: Competitive GEO in AI Answer Engines : SIGIR 2026.