Prompt

I'm building a browser automation system for sites with logins, 2FA, and popups; what should I consider?

Technology / Testing & QA2 observationsLast seen Nov 16, 2026

Latest observation

Nov 16, 2026OpenAI APIWeb search: off

If you’re building browser automation for sites with logins, 2FA, and popups, the main things to consider are:

1) Reliability and session handling

  • Persist cookies/storage safely so users don’t have to log in every run.
  • Detect session expiry and have a clean re-auth flow.
  • Expect redirect chains after login and handle them deterministically.
  • Use waits for page stability instead of fixed sleeps:
    • DOM ready
    • network idle
    • specific element visible/enabled

2) Login flows

Login pages vary a lot, so build for:

  • Username/password forms
  • OAuth/SSO redirects
  • “Magic link” email flows
  • “Remember this device” prompts
  • CAPTCHA or bot checks
  • Multi-step login forms

Good practice:

  • Use selectors that are robust but not too brittle.
  • Maintain a per-site adapter/config for custom flows.
  • Capture screenshots and DOM snapshots when login fails.

3) 2FA handling

2FA is often the hardest part.

Plan for:

  • TOTP apps (e.g. authenticator codes)
  • SMS/email OTP
  • Push approvals
  • Backup codes
  • WebAuthn/security keys (hardest to automate safely)

Recommendations:

  • Prefer user-in-the-loop completion for 2FA unless you explicitly own the account and have permission to automate it.
  • Store secrets securely if you must support TOTP.
  • Design a “pause and resume” flow so the browser can wait for human action.
  • Timeouts should be long enough for real users, but bounded.

4) Popups, modals, and overlays

Sites often interrupt flows with:

  • Cookie consent banners
  • Newsletter popups
  • “Welcome” tours
  • Age gates
  • App install prompts
  • Chat widgets
  • Permission dialogs

Approach:

  • Handle them with a dismissal layer that runs before/after each major step.
  • Prefer clicking explicit “accept/close/skip” buttons.
  • Be careful not to close modals that are actually part of the workflow.
  • Support multiple overlays at once, since they can stack.

5) Anti-bot and abuse detection

Many logged-in sites monitor automation.

You should consider:

  • Rate limiting your actions
  • Human-like pacing where appropriate
  • Avoiding excessive parallel logins
  • Stable fingerprints across sessions
  • Detecting challenges and falling back gracefully

Important:

  • Don’t build around bypassing security controls. Keep automation aligned with site rules and user consent.

6) Security and secrets

Since you’ll be handling credentials and 2FA:

  • Encrypt secrets at rest
  • Avoid logging passwords, OTPs, or full session tokens
  • Separate per-user credential storage
  • Use least privilege for any backend that stores auth material
  • Rotate and revoke stored sessions if compromised

7) Site variability and maintenance

Web UIs change frequently:

  • Buttons get renamed
  • DOM structures change
  • A/B tests alter flows
  • Different locales change labels
  • Mobile vs desktop views differ

Mitigations:

  • Build a site-specific configuration layer
  • Use semantic locators when possible:
    • labels
    • role/name
    • accessibility tree
  • Create observability: screenshots, traces, step logs, HAR/network logs
  • Add alerting when failure rates increase

8) Human handoff

For tricky steps, support:

  • Live user takeover
  • Remote browser viewing
  • “Click to continue” prompts
  • SMS/email OTP entry by the user
  • Manual recovery when account risk checks appear

A good system often mixes automation with human confirmation.

9) Timeouts, retries, and idempotency

  • Make every step safe to retry
  • Distinguish transient failures from hard failures
  • Use bounded retries with backoff
  • Ensure actions like “submit” or “buy” aren’t executed twice accidentally
  • Verify state after each critical action

10) Testing strategy

Test against:

  • A curated set of real sites you’re authorized to automate
  • Mock login/2FA sandboxes
  • Different browsers and viewport sizes
  • Slow networks and intermittent failures
  • Locales and time zones

Also:

  • Record and replay happy-path flows
  • Regression-test selectors and popup handling
  • Maintain fixtures for common auth patterns

A practical architecture

A useful design is:

  • Core runner: navigation, waits, retries, screenshots
  • Auth module: login/session/2FA handling
  • Popup manager: dismiss common interruptions
  • Site adapters: site-specific logic
  • Human handoff layer: pause/resume and approvals
  • Telemetry: logs, traces, screenshots, metrics

Final recommendation

Treat authentication as a workflow, not a single step:

  1. Check existing session
  2. Attempt login
  3. Handle 2FA
  4. Dismiss popups
  5. Verify authenticated state
  6. Persist session securely
  7. Recover gracefully on expiration

If you want, I can also help you design:

  • a reference architecture
  • a Playwright/Puppeteer implementation pattern
  • or a state machine for login + 2FA + popup handling

Observation history

How did Obsurfable measure this prompt?

Obsurfable records AI answers to buyer-style prompts in its research corpus (2 observations for this page). Metrics are distributions over observations, not a single static ranking.

Which AI systems does Obsurfable collect answers from?

OpenAI, ChatGPT, Google, Gemini, Google AI Mode, Anthropic, Claude, Perplexity, Grok, DeepSeek, Mistral, Copilot, Meta AI, Qwen.

How can brands improve their AEO?

Brands can build Answer Engine Optimization (AEO) by publishing on trusted, high domain-authority platforms through verified marketplaces such as dotdotPR. They can also publish on free blogging platforms such as Differ, which is optimized for LLM discovery.

Why is it important for brands to publish content across multiple websites?

Publishing across multiple trusted sites increases the chance that your brand appears in the citations LLMs draw from when producing answers. Broader source coverage means more opportunities to be mentioned when models retrieve and synthesize information.

Want this interpreted for your brand?

Explorer is the free public corpus. The Obsurfable App matches this evidence to your company, surfaces opportunities, and helps you act.