Add your own AI provider keys
AEO, the Suggested Plan brief, Template Studio drafting and AI Enrichment all run on the AI accounts your workspace connects. Ask uses them too, for anyone who has turned it on: Ask a question about your records lists what it sends. You paste a key, MarqIQ proves it works against the provider before storing it encrypted, and you can then restrict every AI feature to those accounts alone. To take your first visibility measurement once a key is in, go on to See whether AI assistants recommend you.
Before you start
Permissions required
Only workspace admins can connect, disconnect or change AI provider settings. Everyone else can open the page and see what is connected. A read-only workspace is refused every change.
Plan required
The AI & Providers settings page is open to any plan with a feature that uses AI accounts — AI Enrichment, AEO, Template Studio or the Suggested Plan brief. The Connect AI engines modal inside AEO needs a plan that includes AEO. A workspace with keys stored but no such feature left can still see them and disconnect them.
- A login to each provider you want to connect. You are usually already signed in, and each key page is one click from the screen.
- Ten minutes for a first-time provider sign-up, which is longer than the connecting itself takes.
The two places you can do this
- AI & Providers, under account settings, is the workspace-level screen. It holds the accounts and the policy, and every AI feature reads the same keys from it.
- Connect AI engines, on the AEO Overview tab, opens the same list in a modal so a first run does not need a detour into settings.
Note
Both write to the same place. A key connected in AEO shows on the settings page and the other way round, so the two screens cannot disagree about what is connected.
Connect a key
Open AI & Providers under account settings.
From AEO, click Connect AI engines on the Overview tab instead — the steps below are the same in both.
Find the engine you want under AI accounts.
Six are listed. Each row carries how to log in and what the provider charges, and Perplexity is badged powers Citations.
Click Open key page beside that engine.
It opens the provider's own key page in a new tab. The addresses are in the table below.
Create a key at the provider and copy it.
Paste it into the field beside that engine.
The field names the provider's own product: Paste your OpenAI platform key, Paste your Google AI Studio key, and so on. A value under 8 characters is refused with "Paste the key from the provider page."
Click Verify & connect.
MarqIQ calls the provider's model list with the key, which spends no tokens, and refuses a key that cannot reach the model AEO measures with.
Check the row now reads Verified. MarqIQ's AI features run on this account.
Inside AEO the modal says Verified and in use for visibility runs. instead, and closing it with Done — continue returns you to the run button.
| Engine | Key page | What the provider charges |
|---|---|---|
| ChatGPT (OpenAI) | platform.openai.com/api-keys | Paid per use — a few cents per visibility run. |
| Gemini (Google) | aistudio.google.com/apikey | Free tier included. If your team is in the UK, EEA or Switzerland, Google's terms allow only its paid tier: use a key from a Google Cloud project with billing turned on. Elsewhere, Google uses what is sent to its free tier to improve its products, and AI Enrichment sends contact details to Gemini with this key. |
| Perplexity | www.perplexity.ai/account/api | Pro includes credits. The one engine that discloses citations. |
| Claude (Anthropic) | console.anthropic.com/settings/keys | Paid per use — a few cents per visibility run. |
| Groq | console.groq.com/keys | Free tier included. |
| OpenRouter | openrouter.ai/settings/keys | Free models included (rate-limited). |
The same six addresses are behind the Open key page links on screen.
What each badge on a row means
| Badge | What it means |
|---|---|
| Your account | You have connected a key for this engine. In the AEO modal the same state reads Connected — your account. |
| Included free | No key of yours, but a key of MarqIQ's can serve this engine. Included engines are off unless your deployment has them switched on, so most workspaces never see this. |
| Included — switched off by your policy | The same, with Use only the AI accounts we own turned on, so MarqIQ's key will not be used. |
| Needs your key | Nothing can reach this engine until you connect one. Every paid engine is always in this state until you do. |
Two cases where the model is not checked
OpenRouter's model is picked fresh on every run, so its key is proved only as a working key and never against a model. And a model list that answers but cannot be read lets the key through, because the key is what was being verified. In both cases a key can be accepted and still fail on a run, so read the first run's health line before you trust it. Perplexity is checked the other way: its model list answers to any key at all, so MarqIQ proves the key with an authenticated read first.
Restrict everything to your own accounts
The Where contact data may go card carries one switch. It decides whether AI answer tracking and the Suggested Plan brief may use an AI account of MarqIQ's where you have not connected one.
Connect at least one account first.
The switch stays disabled until you have, and the server refuses the combination as well — turning it on with nothing connected would leave every AI feature refusing with no way out.
Tick Use only the AI accounts we own.
Wait for Saved. under the card.
- AI Enrichment uses only the accounts connected here whether the switch is on or off, because it sends contact details — name, email, company, job title, country and industry — to each account it uses, and never to an account of MarqIQ's.
- Template Studio drafting also uses only the accounts connected here, whatever the switch says.
- Exact-match lookups such as Gravatar and GitHub are not AI and are unaffected.
- Where any engine is served by an account of MarqIQ's, turning the switch on can reduce coverage: answer tracking then asks only the engines you have connected. The card says so on screen when that applies to you.
Disconnect a key
Click Disconnect on the engine's row.
Read the row after it reloads.
The key is deleted. It is never shown back at any point, not even to you, so a key you want to keep must be kept at the provider.
Take care
Removing your last own account is refused while Use only the AI accounts we own is on, because that state leaves every AI feature with nothing to run on. Turn the switch off first, or connect another engine, then disconnect.
If it did not work
"… rejected the key — copy it again from the key page."
Why: The provider answered 401 or 403. Usually the key was copied short, or it has been revoked since.
Fix: Create a fresh key on the provider's key page and paste the whole value. Nothing is stored until the check passes, so a rejected attempt leaves the row as it was.
"… accepted the key, but the model MarqIQ measures with is not available on it."
Why: The key works, but the project behind it cannot use that model — often a retired model, or a tier that does not include it.
Fix: Enable billing on the key, or use a key from a project that can reach the model the message names. The model is set for the deployment rather than in your workspace, so support is the right route if no key of yours can reach it.
"Paste the key from the provider page."
Why: The field held fewer than 8 characters, so it cannot be a provider key.
Fix: Copy the key again — most providers show it once, at creation, and never again.
"Only workspace admins can change AI provider settings."
Why: Connecting and disconnecting are admin-only, checked against the database rather than the page you are on. Posting from elsewhere is refused the same way.
Fix: Ask a workspace admin to connect the key. You can still open the page and see what is connected.
The switch cannot be turned on, with "Connect at least one AI account above before turning this on".
Why: The policy with nothing connected would leave enrichment with nothing to run on, so that state is refused.
Fix: Connect one engine, then tick the switch.
Disconnecting is refused with a message about this being the only account you own
Why: Use only the AI accounts we own is on and this is your last own key.
Fix: Turn the switch off, or connect a second engine, then disconnect this one.
A key was accepted and fails on every run
Why: The key was checked without a model behind it — OpenRouter, whose model is picked fresh each run, or a model list that answered and could not be read. The key itself was proved; the model was not.
Fix: Read the amber banner above the tiles on the AEO Overview tab — it names the engine and the provider's own reason. Why one AI engine returned nothing, or half your prompts were not asked covers each reason.
Your plan has no AI feature left and keys are still stored
Why: Keys connected on a trial or an earlier plan stay stored. No downgrade clears them, and no other screen can remove them.
Fix: Open AI & Providers. A workspace in this state sees only what is connected, with an amber note naming the plan, and an admin can disconnect each account there.
Next steps
Related articles
Was this article helpful?