Home/ Tools/ AI Answer Absorption Analyser
E.02 · Evidence & Scoring

Measure the six structural dimensions associated with whether AI absorbs your content.

Citation and absorption are different phenomena. Citation means AI lists your source. Absorption means AI draws from your content to shape its generated answer — a far more influential form of engagement. This tool measures the structural dimensions Zhang et al. 2026 found to distinguish high-absorption from low-absorption pages.

ToolE.02 · v2.0
Built forContent creators optimising for AI answers
Time to use~ 3 minutes
OutputScore /100 + selection vs. absorption split
SourceZhang et al. 2026 · preprint (not yet peer-reviewed)
⚠️

Directional guidance. This tool is built on Zhang et al. 2026, a preprint that has not yet been peer-reviewed. Findings are directional — use as guidance, not as hard benchmarks. Next peer-review status check: August 2026.

What is Answer Absorption? ▼
Answer absorption is distinct from citation selection. Citation selection is when an AI system lists a source in its citations. Absorption is when that source actually shapes the words and claims in the AI's generated answer — a more direct and influential form of AI engagement.

This distinction is established in Zhang et al. (2026), 'From Citation Selection to Citation Absorption', a preprint study analysing 21,143 citations across ChatGPT, Google AI Overviews, and Perplexity. The study found that high-influence pages — those whose content was visibly absorbed into AI answers — shared specific structural and content properties: they were significantly longer, had far more headings and paragraphs, and were more likely to contain definitional and comparative language.

This tool operationalises those structural dimensions. It does not predict whether any specific piece of content will be absorbed by any specific AI system at any specific time. It measures the structural properties associated with higher absorption probability.
ABS

AI Answer Absorption Analyser

Measure the structural and content properties that research associates with AI answer absorption — whether AI draws from your content to shape its answers, not just lists it as a source.

Research basis: This tool's scoring model is based primarily on Zhang et al. (2026), 'From Citation Selection to Citation Absorption' — a preprint study of 21,143 citations across ChatGPT, Google AIO, and Perplexity. This paper has not yet completed peer review. Findings are directional and should be treated as current best-available evidence, not settled research. The tool will be updated if the published version materially changes the findings.

Research basis per dimension ▼
Word count, headings, paragraphs (structural dimensions):

Word count: Word count tiers are calibrated from Zhang et al. (2026, preprint — not yet peer-reviewed), which found high-influence pages were on average 11.44× longer than low-influence pages. The specific word count thresholds are the tool's internal calibration.

Headings: Heading count tiers are calibrated from Zhang et al. (2026, preprint — not yet peer-reviewed), which found high-influence pages had 12.50× more headings. Specific tier thresholds are the tool's internal calibration.

Paragraphs: Paragraph count tiers are calibrated from Zhang et al. (2026, preprint — not yet peer-reviewed), which found high-influence pages had 5.69× more paragraphs. Specific tier thresholds are the tool's internal calibration.

A separate preprint study (arXiv:2603.29979, 'Structural Feature Engineering for Generative Engine Optimization', 2026) provides independent corroboration for these structural dimensions. That study tested structurally distinct variants of the same content — identical semantic payload, only structure changed — across six generative engines, and found structural changes alone produced a mean 17.3% improvement in citation rate. This is controlled-experimental evidence that structural features have a measurable effect on citation outcomes, independent of semantic content changes.

Definition sentences:

Definition sentence detection is based on Zhang et al. (2026, preprint — not yet peer-reviewed), which found pages with high definitional content showed approximately 57% higher absorption. This is a page-level finding applied as a document-level dimension — an informed inference, not a directly measured sentence-level effect.

Comparative sentences:

Comparative sentence detection is based on Zhang et al. (2026, preprint — not yet peer-reviewed), which found comparative content was associated with approximately 55% higher absorption. Page-level finding applied as a document-level dimension.

Statistics & numeric evidence:

Statistics presence is associated with approximately 61% higher absorption and code/numeric content with approximately 76.88% higher absorption in Zhang et al. (2026, preprint — not yet peer-reviewed). These are absorption-side measurements: they measure how much a source shapes AI-generated answer language. A separate peer-reviewed study (Aggarwal et al. 2024) found approximately +31% for citation selection probability — this is a different measurement type on a different phenomenon. The two figures are not comparable and are not presented alongside each other in this tool's scoring.

Your Readiness Score will appear here after you click Analyse.
What the research shows

What Zhang et al. 2026 found across 21,143 citations.

R.01

Length is the most visible structural differentiator

High-influence pages — those whose content was visibly absorbed into AI answers — were on average 11.44 times longer than low-influence pages. This doesn't mean longer is always better; it reflects that substantive, well-developed content tends to provide more to draw from.

R.02

Heading density indicates quality to AI systems

High-influence pages had 12.50 times more headings than low-influence pages. Headings appear to function as quality proxies — structured content is parsed differently. The tool measures heading count against word count to surface disproportionate sparseness.

R.03

Definitional and comparative language drives absorption

Content with strong definitional sentences ("X is a process of…") and comparative language ("compared to…", "unlike…", "whereas…") showed measurably higher absorption probability. These patterns appear to signal analytical depth that AI systems prefer when constructing answers.

R.04

Statistics presence is the highest-confidence cross-study finding — an AI citation predictor (Aggarwal et al. 2024, peer-reviewed) and a structural absorption dimension (Zhang et al. 2026, preprint).

Statistics presence is the one dimension corroborated by both Zhang et al. 2026 (preprint) and Aggarwal et al. 2024 (peer-reviewed). Pages containing statistics showed approximately 61% higher absorption and +31% citation probability. The tool labels this as Tier 1 — the most actionable item to fix first.

Research benchmarks

High-influence page properties from Zhang et al. 2026.

11.44×
Length of high-influence vs. low-influence pages
Zhang et al. 2026 · preprint
12.5×
Heading density of high-influence pages
Zhang et al. 2026 · preprint
5.69×
Paragraph density of AI-absorbed content
Zhang et al. 2026 · preprint

All absorption benchmarks are from Zhang et al. 2026 preprint (21,143 citations across ChatGPT, Google AI Overviews, Perplexity). Not yet peer-reviewed. Use directionally.

Measure your absorption dimensions.

Run your content through the Absorption Analyser, then compare with Evidence Density Score for a full picture of AI citation and absorption probability.

Analyse your content ↑ ← Evidence Density Score