What Is Smoke Testing? A Developer's Plain English Guide

Smoke testing is a software testing technique that runs a subset of test cases covering the core functionalities of a new build to verify it is stable enough for further testing (BrowserStack, 2026). The name comes from hardware testing: if you power on a new circuit board and smoke comes out, you stop. You don't run diagnostics on burning hardware.

In software, the equivalent is simpler. A smoke test checks whether the application launches, whether users can authenticate, whether the primary user flow completes without a fatal error. It does not check every screen. It does not validate edge cases. It answers: is this build alive?

Three properties define a real smoke test:

• Narrow scope. Only critical paths, never exhaustive coverage.
• Fast execution. A smoke suite that takes 45 minutes is not a smoke suite. It's a mini-regression that nobody will run consistently.
• Binary result. Pass means proceed to deeper testing. Fail means reject the build and fix it first.

If your smoke test suite takes longer than 10 minutes to run, trim it. Speed is the entire point.

These two terms get swapped constantly, and the confusion causes real process problems.

A smoke test is broad but shallow. It covers many features at a surface level to verify the build is stable. You run it immediately after every new build, before any other testing begins.

A sanity test is narrow but deeper. After a specific bug fix or a targeted code change, a sanity test verifies that the particular area that changed still works correctly. You run it after receiving a build from development when a specific fix needs verification.

The sequence matters: smoke test first, sanity test second. A build that fails a smoke test never reaches sanity testing. Running sanity tests on a broken build is wasted effort.

For mobile apps specifically, the smoke suite typically covers login, core navigation, and one primary transaction per feature area. That's three to eight test flows for most apps. See our AI Smoke Testing for Mobile Apps: A Practical Guide for what a well-scoped mobile smoke suite looks like.

The traditional argument for smoke testing was about saving QA time. The current argument is about CI/CD pipeline integrity.

When you push code multiple times a day and deployments are automated, a broken build can propagate downstream quickly. Without a smoke test gate, a bad deployment might reach staging, trigger a full regression run that takes two hours, and then fail in the first five minutes. You've burned the pipeline and the time.

A smoke test as a pipeline gate stops that. GitHub Actions, Fastlane, and Expo EAS can trigger a smoke suite on every PR merge or every deployment to staging. If the smoke suite fails, the pipeline stops. The team gets notified before anything expensive happens.

Autosana integrates directly with GitHub Actions, Fastlane, and Expo EAS to run automated smoke flows as part of your deployment process. You configure it once, write the smoke flows in plain English, and the agent executes them against your iOS or Android APK or your web app URL on every trigger. Results come back with screenshots at every step so you can see exactly what passed and what failed, without reading through log files.

Teams that add a smoke gate to CI/CD catch critical regressions within minutes of the commit that caused them (QASource, 2026). That's the feedback loop that actually prevents production incidents.

Traditional smoke test automation had a specific failure mode: the tests broke whenever the UI changed. A button moved, a selector changed, an ID got renamed in a sprint, and suddenly the smoke suite that was supposed to protect the build became the thing the team had to fix before they could test anything.

AI smoke testing solves this with two mechanisms working together.

First, natural language test creation. Instead of writing XPath selectors or Appium locators, you write: 'Log in with test@example.com, navigate to the checkout screen, and verify the cart is visible.' The AI agent reads that, identifies the relevant UI elements at runtime, and executes the flow. No selectors to maintain, no script to update when the UI changes.

Second, self-healing test execution. When a UI element moves or a screen layout changes, the agent adapts without a manual update. It uses computer vision and contextual understanding to find the right element in its new position. The test continues. The build verification still happens.

This matters specifically for mobile apps, where UI changes happen constantly across sprint cycles. Autosana's self-healing tests automatically adapt to UI changes so the smoke suite keeps running accurately as the app evolves, without requiring a QA engineer to maintain the scripts.

For teams comparing approaches, the Agentic AI vs Codeless Testing: Key Differences article explains why self-healing AI agents are meaningfully different from older codeless recording tools that still break on UI changes.

Most teams over-scope their smoke suites. They add too many flows, the suite gets slow, and eventually the team starts skipping it. Keep the smoke suite to seven flows or fewer and cover only the paths where failure means the app is unusable.

For a typical mobile app, the smoke suite should cover:

• Authentication. Can a user log in? Can a user log out? If login is broken, nothing else matters.
• App launch and home screen. Does the app open without crashing? Does the primary screen load with expected content?
• Core navigation. Can the user reach the primary sections of the app from the main navigation?
• One primary transaction. For an e-commerce app, that's adding an item to a cart. For a fintech app, that's initiating a transfer. For a subscription app, that's accessing gated content after login.

That's four to five flows. With AI execution, those four flows should complete in under five minutes. If the suite takes longer than that, you've added things that belong in regression testing, not smoke testing.

For specific flow examples, see Mobile App Login Flow Testing With AI and Mobile App Payment Flow Testing AI.

If your smoke suite has any of the following properties, it's not functioning as intended.

The suite runs only once a week. A smoke test that doesn't run on every build is a scheduled check, not a gate. It will miss the regression it was built to catch.

Engineers update the smoke scripts manually after every sprint. This means the suite is built on brittle selectors. It's working against the team, not for them. Every UI change creates maintenance debt.

The suite takes more than 15 minutes. Cut it. A slow smoke suite gets skipped under deadline pressure, which is exactly when it's most needed.

Nobody reviews the results. If smoke test results sit in a log that nobody reads, the suite has zero value. Results need to route to Slack or email automatically so failures are visible immediately.

Autosana addresses the last two directly: the agent sends results via Slack notifications with screenshots at every step, so failures surface immediately to whoever needs to act on them.