Methodology

Peopling Earth draws on several interlocking lines of scientific evidence:

Ancient DNA (aDNA) Genomic analysis of ancient human remains is the most direct way to detect prehistoric migrations. aDNA can identify population replacements, admixture events, and genetic relationships invisible to archaeology or linguistics. All aDNA findings cited on this site derive from peer-reviewed studies published in journals such as *Nature*, *Science*, *Current Biology*, and *PNAS*.

Modern population genetics Genotype data from living populations—haplogroup distributions, admixture proportions, and principal-components analyses—corroborate or refine ancient DNA conclusions.

Archaeology Material culture—stone tools, pottery traditions, burial practices, settlements—provides a spatial and temporal framework that genetics alone cannot supply. Archaeological cultures described on this site follow mainstream academic classification.

Linguistics The comparative method in historical linguistics reconstructs proto-languages and proposes homelands for language families. When linguistic spreads align with genetic migrations, confidence in both increases.

Palaeoclimatology Ice-core, pollen, and sediment records reconstruct the climatic conditions that opened or closed migration corridors. Climate proxies are cited to explain *why* particular migrations occurred when they did.

Written records and oral traditions Historical documents and oral traditions can corroborate events detected through other methods, but are evaluated critically for bias and transmission reliability.

Longer textual content on Peopling Earth—entity summaries, blog posts, and suggested reading—is generated with the assistance of large language models (currently Grok via xAI, with GPT-4 and Gemini as fallbacks). AI generation is used only for prose drafting; all claims are grounded in real academic consensus.

How generation works Each content request uses a structured prompt that specifies: (1) the entity or topic, (2) relevant date ranges and geographic scope, (3) the evidence types to prioritize, and (4) explicit instructions to signal uncertainty, avoid over-claiming, and flag contested interpretations. The model is instructed to produce Markdown with clear section headings.

AI content is never treated as a primary source If a generated summary conflicts with the peer-reviewed literature, the literature takes precedence and the summary is flagged for regeneration. Readers who spot factual errors are encouraged to submit a correction.

The Dr. Amara Vey research assistant The interactive Q&A assistant uses the same model infrastructure but is explicitly designed to present multiple scholarly perspectives, signal uncertainty, and provide links to further reading. Dr. Amara's persona is a pedagogical frame—all answers reflect real scientific debate.

Human prehistory involves large amounts of genuine uncertainty. Peopling Earth uses consistent conventions to communicate this:

• Date ranges with BCE / CE are given when known; "circa" (c.) notation indicates estimates.
• "Proposed homeland" language for language families and migrations indicates hypotheses supported by current evidence, not established facts.
• Summary entries for some topics are intentionally concise and will be expanded over time. All published summaries are accurate but may not yet reflect the full complexity of the topic.
• Contested theories are flagged with language such as "debated," "not universally accepted," or "some researchers argue." Peopling Earth does not adjudicate unresolved debates.

Maps and population descriptions on Peopling Earth are necessarily approximate:

Genetic clusters are not ethnic groups Ancestry components identified through admixture analysis are mathematical constructs that reflect population history, not discrete biological categories. Labels like "Yamnaya ancestry" describe a statistical signal, not a people who called themselves Yamnaya.

Modern ethnic identities are not ancient identities When this site uses terms like "Celtic" or "Bantu," it refers to archaeological or linguistic traditions, not modern nationalist categories. We avoid implying that ancient and modern groups sharing a name are identical.

Geographic ranges are approximate Population distributions shown on maps represent scholarly best estimates at particular time slices. Actual distributions were continuous gradients, not discrete polygons.

Terminology standards Peopling Earth follows the terminology guidelines of the American Anthropological Association and the conventions of the journals cited in its sources. Readers who identify terminology that is outdated, disrespectful, or contested are encouraged to submit a correction.

Peopling Earth is an independent educational project. Its editorial process involves:

1. Source grounding. All entity data is seeded from peer-reviewed literature. References to specific studies are noted in content where space permits.
2. AI drafting. Longer prose is drafted by AI with structured prompts that enforce citation discipline.
3. Human review. All published content is reviewed before release. Entries that require more detailed treatment are expanded on an ongoing basis.
4. Reader corrections. Readers may submit corrections through the corrections page. All submissions are reviewed; significant corrections are noted on the affected page.
5. Continuous improvement. Entity summaries are enriched on an ongoing basis. News items are filtered by an AI relevance classifier before publication.