Inbox test: see where your email lands before you send, and what an agent checks first

Inbox Test

An inbox test seeds a message to real mailboxes across providers and reports which folder it lands in, so you can see placement before sending to your full list.

When an AI agent sends the campaign over MCP, it can draft, review and test the email without anyone watching, and it tends to read a delivered status as proof of placement. An AI agent can review the email and send the test messages, but the folder is confirmed by a seed test or by a person opening each test mailbox, before anyone approves the full send.

A seed test is a test of your setup only. It reads your domain and content, never your relationship with real recipients.

What changes when an AI agent runs an inbox test

An agent working through the Nitrosend MCP server or a coding assistant such as Claude can draft the campaign, review it and fire the test sends in one sitting. The risk is that it reads "accepted" or "delivered" as "inbox" and moves on.

George Hartley, Nitrosend's Co-founder, makes a second point in the deliverability reference of his Email Marketing Bible: a warm-up score measures warm-up mail, not campaigns. A mailbox can score 99 while its campaigns land in spam, so he confirms with a seed test before launch and monthly after that.

His send gate for agents puts the pre-send checks first. Seed tests to the person's own addresses can go out without a separate yes, but any send to more than one recipient waits for the words "send it". Reading the folder stays with a person or a seed test.

What it tells you

An inbox test shows how much of your mail lands in inbox, promotions or spam at Gmail, Outlook, Yahoo and others. That is information your own logs can't give you, and it's the check to run before a large or important campaign. It is a direct measurement of inbox placement, which is why it sits at the center of the email deliverability workflow rather than at the end of it. The metric itself, and how it is reported, is covered in inbox placement rate.

The result is also the input to your deliverability rate that logs cannot supply, since logs report acceptance, not folder. Acceptance on its own is the subject of email delivery. If the test comes back poor, what email deliverability is lays out the four layers to work through, and the email deliverability best practices page covers the habits that stop the result drifting again.

Two common starting points for a poor result have their own pages: sender score for sending reputation, and new domain emails going to spam for a domain with little sending history. Provider rule changes that can move a result are tracked in email deliverability news.

When to run one

Run an inbox test at these points:

• Before big campaigns.
• Before a launch, and monthly after that.
• After any authentication or infrastructure change.
• When troubleshooting a drop in placement.

Pair it with a spam test email for the content and authentication score behind the placement result. The mechanics of the send itself are in how to send a test email.

George runs the same pre-send check every time. His list:

• Resolve wrapped button and link URLs.
• Check the spam score.
• Check the ratio of image to text.
• Check how it renders in dark mode.
• Make a test send to a seed address.

The deliverability chapter of his Bible diagnoses placement with a step-by-step triage covering authentication, reputation, content and the AI-mediated inbox, where Gemini summaries sit in the path and open rate is noise.

One content trap George flags is the image-only email. It gives Gmail, Yahoo and Outlook no text to classify, so he adds a line of live text and alt text that says what the email is about.

Remembering to run the test before a big send is a job an agent can carry. Asked to send a campaign, it can run the review and the test send first, then stop until a person has read the folders.

How Nitrosend fits into an inbox test

Nitrosend's tools cover the sender's side of an inbox test: the domain, the content and the test copies. The folder itself is read by a person opening each test mailbox, or by a seed test.

• nitro_manage_domains diagnoses the sending domain, so an authentication gap shows up before any test copy goes out.
• nitro_review_delivery gives a read-only spam analysis and previews of the draft. It never sends.
• nitro_configure_account holds the test recipients, which is where the seed addresses live.
• nitro_send_test_message sends one real copy to those addresses and never reaches the audience.

It is worth running against transactional email templates too, not just campaigns, because a password reset in spam is the failure nobody reports. A template the app will later send through the email API or the transactional email API can get its test send over MCP first.

Test sends make this cheap to do early: they use your normal email credits and daily allowance, and the Free plan's 5,000 starter emails cover setup, campaign and flow tests. The daily recipient tiers are on the pricing page.

After launch, Google Postmaster Tools shows how Gmail sees the domain over time, which a one-off seed test cannot.

Worked example: an agent checks a launch email before it goes out

The prompt: "Check where our product launch email lands at Gmail, Outlook and Yahoo before we send it to the list."

Before. Left unguided, an agent sends one test, sees "delivered" in the response, reports the launch email as fine and schedules it for the whole list. Nobody has looked at a folder.

After. With the Nitrosend MCP tools, the agent works through a fixed order:

1. nitro_get_status confirms the account is ready and which sender identity the launch email will use.
2. nitro_manage_domains runs a diagnose on the launch domain before any seed address gets a copy.
3. nitro_review_delivery reads the launch campaign draft without sending it, and its spam analysis flags the hero as a single image with no live text.
4. nitro_compose_campaign patches the draft with a line of live text and alt text on the hero, leaving the audience and schedule as they were.
5. nitro_configure_account reads back the test recipients: the person's own Gmail, Outlook and Yahoo addresses.
6. nitro_send_test_message mails one copy of the launch campaign to those three addresses. Nobody on the list gets it.
7. The person opens each mailbox and reports the folders. The Outlook copy is in junk, so the launch is held.
8. After another content fix and a retest that reaches the Outlook inbox, the person says "send it", and only then does nitro_control_delivery approve the campaign.
9. nitro_get_insights then reports the launch's email analytics with trends, benchmarks and recommendations.

The report back is one line per provider: Gmail inbox, Outlook junk then inbox after the retest, Yahoo inbox. The launch stays held until a person approves it.

Where an agent's inbox test goes wrong

• Accepted or delivered read as inbox. A send response says the receiving server took the message, not whether it went to the inbox, promotions or junk.
• One Gmail address treated as every provider. Gmail, Outlook and Yahoo filter differently, so a test copy in one Gmail inbox says nothing about an Outlook junk folder.
• A warm-up score taken as campaign placement. The score measures warm-up mail, so a high number can sit next to campaigns that land in spam.
• The test sent after the campaign was already approved. A junk result that arrives after approval cannot hold a launch that is already scheduled.
• A clean seed result read as proof the real audience will engage. Seeds have no engagement history, so they read the setup, not how real recipients treat the mail.

FAQ

Go deeper