CNN: A Case Study in Presidential Coverage

Living investigation — preliminary snapshot: This draft currently contains 2,542 CNN URL/headline records that meet the operational rule: A name-matched record explicitly names by surname the person who was the sitting U.S. president when the article was published. This is not every article “about” a president; semantic aboutness is broader and requires a separate review dataset.

Last refreshed: July 20, 2026 at 02:06 UTC. The historical crawl is incomplete, so these figures are operational checkpoint data—not comparative findings about CNN.

CNN census in progress: January 20, 2013–present

Frozen database snapshot: Seventeen non-overlapping administration-year workers are inspecting the 638,562-page study-period inventory derived from CNN’s current and legacy first-party article sitemaps.

Term-aligned headline tone: preliminary growing crawl view

Important: these lines begin after a 25-headline cumulative burn-in and stop at each administration’s current crawl-completion-equivalent frontier. No artificial 0% or 100% starting value is inserted. They extend rightward only as more pages are inspected. This is a nonrepresentative preliminary discovery sample, not a completed comparison of CNN coverage. Obama II and Biden are blue; Trump I and Trump II are red. Dashed lines distinguish Biden and Trump II. The cards preserve the certified continuously completed-span sample separately; it remains zero until an uninterrupted crawl window is finished.

Preliminary cumulative negative headline share

Preliminary cumulative neutral headline share

Preliminary cumulative positive headline share

Preliminary cumulative headline tone score (−100 to +100)

Visual checkpoint only. The x-position is a crawl-completion-equivalent frontier across the four-year frame, not a claim that the current sample is chronologically complete. Apparent differences can shrink, reverse or disappear.

Current merged sitting-president name-match corpus: 2,542 records

Reader scale: the tone score is VADER compound multiplied by 100. −100 is maximally negative wording, 0 is lexically neutral, and +100 is maximally positive wording.

Current positive, neutral and negative headline shares

VADER classifies headline wording as negative at compound ≤ −0.05, neutral between −0.05 and +0.05, and positive at ≥ +0.05. Lexical sentiment does not measure factual accuracy, fairness, evidentiary support, newsworthiness, intent, or political bias.

Current tone score with 95% confidence intervals

Monthly checkpoint table

Content types

Provisional author lexical screening

Only author-administration groups with at least 30 name-matched headlines are eligible. This table is lexical screening and does not characterize an author as biased.

Crawl completion

The frozen CNN-published sitemap inventory contains 709,150 URL records across current and legacy first-party sources; 638,562 route to the study period. Sitemap coverage is not proof of every page CNN ever published. Removed, deindexed or orphaned pages may be absent. No administration has a certified continuous crawl frontier in this snapshot.

Methodology and limitations

• Inventory: CNN’s declared current article sitemap hierarchy plus its frozen legacy article sitemap index, with raw source responses archived and hashed.
• Boundaries: presidential terms change at noon Eastern on January 20, 2017, 2021 and 2025; page publication timestamps, not URL dates or modification dates, determine the incumbent.
• Evidence: every retrieval retains status, errors, redirects, retrieval time and content hash; qualifying HTML is archived locally.
• Deduplication: URL records, headline-text counts, canonical aliases, redirects and identical content hashes remain distinct measures. “URL/headline records” is used unless headline text is explicitly deduplicated.
• Aboutness: surname matching is reproducible but narrower than semantic coverage about a president.
• Uncertainty: confidence intervals describe lexical-score uncertainty only; they do not establish causal differences or political bias.

Evidence snapshot SHA-256: 93b21f46fdaeb2438d9235b0f9617a22b876290ca6e5e0148681e2bcad5bc641