• Databases: Which databases to use. Their counts appear in the Citations column in the order you list them, separated by |. Hover a cell to see which database each number came from.
  • Default: crossref, semanticscholar
  • For physics papers you might prefer inspire, crossref, semanticscholar

• Fetch tallies for new items: On by default. Turn it off to stop counts being fetched as you add items.
• Automatic updates: Off by default. Set it to refresh missing and outdated counts in My Library the next time Zotero starts.
• Consider citations outdated after: How old a count can be before it is considered outdated — 3, 6, 12, or 24 months. The default is 6. This applies both to startup updates and to Tools → Retally outdated item citations.

• Colors: On by default. Each database's count gets its own color when more than one database is shown; turn it off to display every count in the default color.

• Semantic Scholar API key (optional, but recommended): Enter a key under Settings → Citation Tally → API keys. Without one, requests share Semantic Scholar's anonymous pool with every other client using it, so lookups are slower and less reliable.
  • Request a key at semanticscholar.org/product/api.
  • Zotero stores the key unencrypted in its local preferences. Citation Tally sends it only to Semantic Scholar.
  • If Semantic Scholar rejects the key twice in a row, the plugin stops using it, carries on anonymously, and tries it again after a cooldown. A single rejection triggers one confirming request rather than pausing the key.

• No citation data found: Citation Tally looks items up by DOI or arXiv ID only, not by title, ISBN, PMID, or a database's own record ID. Journal articles and conference papers usually carry a DOI; web pages, theses, and datasets often carry neither identifier.
• Updates are slow: Each database paces its own requests and the plugin backs off further when a server throttles it, so a large update can take a while. See Rate Limiting and Retries.
• Semantic Scholar has been turned off: If Zotero's plugin runtime doesn't provide the web APIs the Semantic Scholar client needs, Citation Tally disables that database, shows a notice, and carries on with the other databases you have configured.
• Network issues: Ensure Zotero has internet access and your firewall isn't blocking requests to academic databases.

Retries

Startup updates and Tools → Retally outdated item citations both scan My Library, and both back off when a database comes up empty: 7 days after the first failure, then 30, then 90, then 180 days for every attempt after that. API errors hit during a scan follow the same schedule.

Rate Limiting

Crossref and INSPIRE each start at one request per second, and are throttled independently. A rate-limit error multiplies the delay by 1.5, up to ten times the base; each success eases it back by 0.9, never below the base.

Semantic Scholar runs its own scheduler: at least 1 second between requests made with an API key, and at least 3 seconds without one. Transient failures back off exponentially with full jitter, never sooner than the server's Retry-After.

Other Behavior

• If an item has no usable identifier for a database, a library scan skips that item–database pair for the rest of the Zotero session. Selected-item updates do not consult this cache.
• A library scan does not start updating items if Zotero is already offline, and stops before the next item if Zotero goes offline during the run.
• Retry records for deleted items are cleared out shortly after startup and every 30 days. Counts already written to the Extra field are left alone.

Supported Identifiers

Crossref needs a DOI. INSPIRE and Semantic Scholar can also use an arXiv ID.

Citation Tally reads the DOI field, then looks for an arXiv ID in Archive ID, Report Number, Extra, URL, and Call Number, in that order.