Which API keys you need, and how they stay safe

Citation tracking is the only part of CiteTrack that uses AI API keys, and you only need a key for the engines you actually want to track. One is enough to begin. Crawler analytics, the part that logs which AI bots hit your site, needs no key at all.

How many keys you really need

Start with one, most people pick ChatGPT or Perplexity, add a prompt, and run it. Add Gemini or Claude later when you want a fuller picture. There is no requirement to wire up all four, and nothing breaks if you only ever use one.

Where to get each key

• OpenAI (ChatGPT), platform.openai.com → API keys → create a key, add a little credit.
• Perplexity, the API settings in your Perplexity account, where you generate a Sonar key.
• Google (Gemini): Google AI Studio issues a free key with a generous daily quota.
• Anthropic (Claude), console.anthropic.com → API keys.

Paste a key into Citations, click Test, and you are done. The setup wizard walks you through it the first time.

How your keys are stored

Keys are encrypted in a local AES-GCM vault and never leave your server except to talk to the provider they belong to. There is no CiteTrack cloud in the middle, your WordPress site calls OpenAI, Perplexity, Google or Anthropic directly, and the answer comes straight back. We never see your keys, your prompts, or your results.

How model and spend control works

CiteTrack tracks provider, model, token usage, search/tool usage and estimated cost in the local cost ledger when that metadata is available. If you select a newer model that is not in the pricing catalogue yet, the ledger marks it as unpriced instead of guessing. You can still run it; the plugin simply keeps the accounting honest.

Whether this is allowed

Yes. CiteTrack uses each provider’s official, paid API and its sanctioned web-search tool. It does not scrape the consumer chat apps or automate logged-in sessions, so there is nothing against any provider’s terms and no risk of a key being flagged for misuse.

← All articles