API Keys (BYOK)

CiteMetrix uses your own API keys to call AI platforms. This is sometimes called BYOK — Bring Your Own Keys. It’s an unusual pricing model in this category, and worth understanding before you set up your first key.

This article covers the concepts: what BYOK means, what each provider’s key actually powers in CiteMetrix, how keys are stored, and the cost picture. For the step-by-step procedure to obtain each key, see the dedicated API Key Setup Guide — that page has up-to-date instructions for every supported provider, including which buttons to click in each provider’s console.

What BYOK means

Three things, in order of impact:

• You pay AI providers directly at their list pricing. No middleman markup. CiteMetrix’s subscription covers the platform; the underlying AI calls are your costs, billed to your card on file with each provider.
• You own the credentials. If you ever leave CiteMetrix, your keys keep working in whatever else you connect them to. There’s no platform lock-in on the AI access side.
• Your scan results reflect each platform’s current responses to your account. Platforms occasionally personalize, and your account’s API responses are the closest thing available to “what AI says about you.”

The trade-off: you have to set up an account with each AI provider you want CiteMetrix to scan. The Setup Guide walks you through that for each one.

What each key powers

Minimum viable setup

Two keys are required on every account: Anthropic (Claude) and Perplexity. They power CiteMetrix’s core citation, sentiment, and accuracy features, so scans won’t run without them. The other seven platforms are optional — add them as you supply their keys. The most useful next additions are OpenAI (one key covers both ChatGPT and Copilot) and Google AI Studio (Gemini), then Grok, DeepSeek, and Mistral for breadth or specific regional reasons.

Don’t try to set up every key in one sitting. Start with the two required keys, run your first scans, see what kind of data you’re getting, and add platforms as you find gaps.

Where to get each key

→ API Key Setup Guide is the procedural reference. It’s organized by provider, walks through account creation, billing setup (most providers require a payment method before generating a key), and shows where to copy each key into CiteMetrix’s settings.

The Setup Guide stays current with each provider’s UI as those change; this docs article doesn’t try to keep up with every minor “the button moved” change.

How keys are stored

CiteMetrix takes API key handling seriously because the alternative is bad:

• Keys are encrypted at rest — account keys under the account, domain-level keys in a per-domain store.
• Keys are decrypted only when needed for an outbound API call. They never appear in logs, debug output, or the dashboard UI.
• Keys never leave your CiteMetrix install. Calls go directly from your CiteMetrix to the AI provider — there’s no third-party proxy in between.
• If you suspect a key has been compromised, revoke it at the provider’s console and generate a new one. CiteMetrix is happy with rotation; nothing breaks except the brief gap until you save the new key.

Where keys live: per domain, then per account

CiteMetrix resolves keys in two layers — which is what lets an agency bill each client’s AI spend separately while everyone else just sets keys once:

• Account-level keys — the default. Set them once under Settings → API Keys, and every domain on the account uses them. This is all most customers ever need.
• Domain-level keys — an optional override for a specific domain (= a specific client). Add a key to a domain and that domain’s scans use it instead of the account key, so provider cost and usage are tracked per client. This is how agencies keep each client’s AI billing separate.

Resolution order, every scan: CiteMetrix uses the domain’s own key for that provider if one is set; otherwise it falls back to the account key. There is no per-user key — team members share the account’s keys rather than entering their own — and no platform or admin fallback below the account. If neither a domain-level nor an account-level key exists for a provider, that platform is simply skipped for that domain’s scan.

Adobe Analytics has a deliberate exception via an admin-controlled agency-credentials toggle (off by default), which lets one set of Adobe credentials serve every domain on an account. This was added because Adobe relationships are typically held by the agency, not the individual operator.

Validation

When you save a key, CiteMetrix tests most providers with a cheap API call to confirm the key works. A wrong key gets caught immediately at save time.

Adobe Analytics is the exception — Adobe needs the full credential set together (Client ID + Secret + Org + Company + Report Suite) to validate, and CiteMetrix lets you save fields one at a time. So Adobe validation happens lazily on first use rather than at save. If your Adobe credentials are wrong, you’ll see the error the first time CiteMetrix tries to fetch Adobe data, not when you saved them.

Quota banner

If a key hits a quota cap or fails authentication during a scan, CiteMetrix surfaces a quota banner in the dashboard with a direct link to that provider’s billing page (for Anthropic that’s console.anthropic.com/settings/billing, for OpenAI it’s platform.openai.com/usage, etc.). The banner is informational — your other configured providers keep working. It clears automatically when the next scan against that key succeeds.

A note on cost

The first month with CiteMetrix usually surprises people in one direction or the other:

• You pay less than expected because per-scan costs are small. A single domain with 10 monitored queries across all nine platforms runs about $7/month scanned weekly, or roughly $50/month scanned daily. The measured anchor from real billing: OpenAI (ChatGPT and Copilot) is about $0.03 per query-scan, and the cheapest platforms (DeepSeek, Mistral) are under a third of a cent.
• You pay more than expected because you have many queries, broad platform coverage, and frequent daily scans. A heavier pattern — 25+ queries, all platforms, daily — runs roughly $100–$150/month. The biggest levers are scan frequency and how many platforms you enable, not the per-call price.

CiteMetrix has a built-in Spend Monitor (Settings → Spend Monitor) that you can configure to circuit-break automatically if Anthropic spend exceeds a daily threshold. Enabling it is recommended, especially while you’re getting a feel for your steady-state cost. The monitor uses a daily threshold you set; if exceeded, background AI work pauses until midnight UTC, then resumes. A tripped breaker costs you a few hours of staleness; an unbounded runaway costs real money.

What CiteMetrix doesn’t charge for separately

• Citation scan API costs are passed through to your AI provider keys. CiteMetrix doesn’t add a markup or a per-call fee.
• Hallucination detection runs on your Anthropic key (Haiku model, very cheap per call). Same passthrough.
• The Analyst runs on your Anthropic key (Sonnet model when reasoning is needed). Each Analyst conversation typically costs $0.05–$0.20 in Anthropic spend.
• ModelScore calculation is computational, not AI-driven — it costs nothing per run.

Next steps

• New to CiteMetrix → start with Getting Started, which walks through first-key setup at a basic level
• Ready to set up keys → use the API Key Setup Guide for the per-provider procedure
• For non-AI providers (GSC, GA4) → Google Integrations
• Before your first heavy scan day → configure the Spend Monitor in Settings