Inside an entry

Each Sourcipedia article is one entry: a summary written from its sources, numbered citations in the text, and the bibliography behind them. This page covers what an entry shows, how it is written, and how it changes.

01Anatomy of an entry

An entry page at /article/:id has these parts, in this order. Parts an entry doesn't have are left out.

Header

The main concept the entry belongs to sits above the title. Clicking it searches for that concept, which lists entries tagged with it. Under the title, a line says when the entry was written, for example "Synthesised 3 days ago". Some entries also show a quality score and a version number. The version is the format the entry was written in, not a count of edits.

Summary

The body of the entry: a short summary written from the entry's sources. Numbered markers in the text are citations. Hover over one, or move focus to it with the keyboard, to see the source's title and website with a Visit source link. When several markers sit together, the pop-up has an arrow to step through them. Each number matches a line in the bibliography.

Concepts

Tags for the topics the entry touches. Concepts are what tie entries together: the related list at the foot of the page and searches by topic name both rely on them.

Key facts, claims and other sections

Some entries carry extra sections: Key facts, Key quotes, Claims and Methodology. Each can be shown or hidden, and longer ones start hidden. Some claims show a confidence percentage. Citation markers in these sections work the same way as in the summary.

Bibliography

Every source the entry cites, numbered. Each line shows the source's title, linked to the original page, and the website it comes from. Sourcipedia links out to its sources. It doesn't host copies of them.

Related entries

Up to eight other entries in the same main concept, newest first.

02How an entry is written

Entries are written by the research engine behind ResearchAnything, running inside Sourcipedia. Nobody writes or edits them by hand. Each topic goes through the same steps:

1. A topic comes off the research queue. See Where topics come from.
2. Sources are gathered. The engine first checks the sources already cited across Sourcipedia, matching on the topic's key words. If those cover the topic, no web search runs. Otherwise it searches the web through You.com, using the full wording of the topic, and searches the scholarly databases Semantic Scholar, OpenAlex, Crossref, PubMed, arXiv and Europe PMC. When the existing sources already cover a good part of the topic, the web search is steered away from the sites they come from.
3. Sources are screened. Duplicates and thin or low-quality pages are dropped. The rest are scored for relevance and authority, and the strongest are read in depth.
4. Duplicates are caught. If the result matches an existing entry, its sources are added to that entry and no new page is made.
5. The entry is written. A model pulls individual claims out of the sources and writes the summary against them, citing each source. A second pass grades the draft for coverage, citations and statements the claims don't support, and a low grade triggers one rewrite. The entry is then tagged with concepts and linked to related entries.

03Where topics come from

Sourcipedia chooses its own topics. The research queue is filled from five places:

• A starter list of specific research questions across science, technology, health, society, the humanities, the arts and mathematics.
• Topics a model proposes, weighted toward areas the encyclopedia covers thinly.
• Concepts that existing entries mention but that have no entry of their own.
• Titles of widely read Wikipedia articles. Only the title is used, as a topic to research.
• Searches on sourcipedia.com that found nothing. These go ahead of routine topics in the queue.

Topics are checked before they join the queue. Typos, noise and attempts to inject instructions are turned away, and topics already in the queue are skipped.

04How entries change

Entries change in two ways:

• New sources are added. When later research on a topic comes out as a near-duplicate of an existing entry, its sources are added to that entry's bibliography instead of starting a second page. Sources already listed are not repeated.
• Duplicates are merged. A background process looks for entries with the same name. When it finds some, a model picks the best one to keep and may revise its title and summary using the others. Their sources are combined into the entry that remains, and the rest are removed.

The page shows when an entry was first written. It doesn't show a last-updated date or a history of changes.

05Permanent links

Each entry has its own address at /article/:id. Adding sources to an entry doesn't change it.

06Quality checks

Sourcipedia checks its work at several points:

• Topics are screened before research starts.
• Sources are deduplicated, screened for quality and ranked before anything is written.
• The summary is written against claims pulled from the sources, then graded, and rewritten once if it scores low.
• Near-duplicate results are folded into existing entries, and entries with the same name are merged.

These are checks, not guarantees. Entries are published without human review, and a model can still get things wrong. The sources are listed under every entry so you can check.

07What an entry is not

• Not reviewed by people. Nobody reads each entry before it is published.
• Not a primary source. Cite the sources in the bibliography, not the entry.
• Not editable. The public site is read-only. Entries change through new research and merging.
• Not an opinion piece. An entry summarises what its sources say.

08Where to go next

• How search finds entries → Search
• The engine that researches each entry → ResearchAnything overview
• How ResearchAnything formats citations in its own reports → Citations