Bardeen Playbook Not Executing the AI Step: Fix Guide

May 18, 2026 · Editorial Team · 6 min read · bardeen troubleshooting error-fix

You built a Bardeen playbook to scrape LinkedIn profiles from a Sales Navigator search, pass each profile summary through Bardeen's AI Prompter action to generate a personalized outreach angle, and append the results to a Google Sheet. The scraping step works fine, the LinkedIn data lands in Bardeen's output panel, but the AI Prompter step either produces a blank output or the step appears in the playbook run log as "Skipped." Your Google Sheet columns for the AI-generated content are all empty. You've run the playbook four times. Same result every time. Here's what's actually going on.

What this error actually means

Bardeen's AI Prompter and AI steps are built on top of a cloud AI call that Bardeen orchestrates on your behalf, using your connected AI provider (typically OpenAI or Bardeen's own model access). When the step shows as "Skipped" or produces empty output, it's almost always because the step received input that didn't match its expected format, the AI API call failed silently, or Bardeen's extension lost context between the scraping step and the AI step due to a page navigation event or tab focus change.

Unlike n8n or Make.com where you can inspect the raw API calls, Bardeen is a browser extension that runs its AI steps partly inside your browser context. This means page state, tab activity, and Chrome's resource limits all affect whether the AI step actually executes.

Quick fix (when you need it working in 60 seconds)

Open Bardeen's sidebar (click the Bardeen icon in Chrome), go to "Runs," and click the most recent failed playbook run.
Expand each step and look for the AI step. Check whether it shows "Skipped," "Error," or "Empty output."
If "Skipped": the input field the AI step was mapped to is empty. Go back and check what the scraping step actually returned. Click the scraper step output and verify the field your AI prompt references (for example, {{profile_summary}}) has actual content.
If "Empty output": click the AI step to expand its config, check the prompt field, and manually paste a sample input value where {{profile_summary}} would be. Click "Test step" if that option is available.
If the manual test produces output, the issue is that the dynamic field reference resolves to empty at runtime. Reconnect the field mapping from the scraping step's output to the AI step's input field.

Why this happens

The most frequent cause is a field mapping that points to a scraper output field that doesn't exist in every scraped item. LinkedIn's profile structure varies: some profiles have a detailed "About" section, others don't. If your AI prompt is built around {{about_section}} and 60 percent of the profiles don't have an About section, those records return an empty string for that field. Bardeen's AI Prompter receives an empty prompt input and either skips the step or returns a blank response, depending on the Bardeen version.

Another common cause is Bardeen's AI step timeout at the browser extension level. Bardeen runs inside Chrome, and Chrome imposes resource constraints on background extension scripts. If you're running a large batch playbook (50+ records), Chrome may throttle or suspend the extension's background worker mid-run. The AI step, which requires an active network connection and a response from an external API, gets no response before Chrome's background task timeout fires. The step records as completed but with no output.

Bardeen's AI Prompter also has a character limit on the input it will process per call. As of Bardeen's 2025-2026 versions, inputs over approximately 4,000 characters are truncated or rejected without an obvious error. LinkedIn profiles with long employment histories can easily exceed this limit when concatenated into a single prompt input.

Finally, if you haven't connected your OpenAI account to Bardeen (or your Bardeen AI credits are exhausted), the AI step will silently fail on free-tier accounts. Bardeen provides a limited number of AI action credits per month, and once they're gone, AI steps stop executing without a clear error message.

Permanent fix

Add a conditional check before your AI step. In the playbook editor, add a "Conditional" step that checks whether the input field is not empty: IF LENGTH({{profile_summary}}) > 20. On the true branch, run the AI step. On the false branch, set a default value or skip to the next record.
Consolidate your field references. Instead of mapping multiple scraped fields separately to the AI prompt, combine them in a "Transform" or "Format Text" step first. Build a single {{combined_profile_text}} field from the available fields (name, title, company, headline), then pass that single field to the AI step. This way, every record has a non-empty input even if individual fields are blank.
Check your Bardeen AI credits. Open Bardeen's settings (the gear icon in the sidebar), go to "AI Usage," and confirm you have remaining credits. If you're on the free plan, upgrade to Bardeen Pro or connect your own OpenAI API key in the settings. Connecting your own OpenAI key bypasses Bardeen's credit limit entirely.
Reduce batch size. Don't run batch playbooks on more than 20-25 records at a time. For larger lists, break the playbook into chunks or use Bardeen's "Schedule" feature to run smaller batches over time. This prevents Chrome from throttling the background worker.
Keep the Bardeen sidebar open and the Chrome window in focus while the playbook runs. Chrome deprioritizes background tasks for extensions when the window is not in focus or when the tab is hidden. Running Bardeen with the sidebar open and the window active prevents this throttling.
Trim your AI prompt input. Before passing scraped text to the AI step, use a "Slice Text" transform to limit it to a maximum of 2,000 characters. This keeps the input within Bardeen's reliable processing range and reduces API call latency.
Use Bardeen's "Debug run" mode (available in the right-click context menu on a playbook). Debug mode runs the playbook one step at a time and shows you the exact input and output of each step, including the AI step. This is the fastest way to see whether the AI step is receiving your intended input.
If you need to process AI steps on large volumes reliably, consider building the AI call into a separate automation (Make.com, Zapier) that Bardeen triggers via a webhook. Bardeen handles the scraping, sends the data to Make or Zapier via an HTTP step, and the more reliable automation platform handles the AI processing.

Prevention

Always test your playbook on a single record before running a batch. Use Bardeen's "Run on current item" option to process just one LinkedIn profile and verify the AI step output is what you expect. Catching an empty field mapping on record 1 saves you from discovering it on record 50.

Check your AI provider connection monthly. Go to Bardeen's settings and confirm your OpenAI or Anthropic connection still shows as "Connected." OAuth tokens for these services can expire or be revoked if you rotate API keys in your provider dashboard without updating Bardeen's connection.

Review your scraper's output fields before building the AI step. Open the scraper step in edit mode and use the "Test scraper" function to see exactly what fields and values it returns for a typical record. Design your AI prompt only around fields that are reliably present across your target page type.

When the fix doesn't work

If the AI step still skips even with a known-good non-empty input and valid credits, it may be a bug in your specific Bardeen version. Check the Bardeen changelog at bardeen.ai/changelog and compare your installed extension version (visible in Bardeen's settings) against the latest release. Update the extension if you're behind.

For persistent AI step failures that don't respond to the above fixes, contact Bardeen support via the in-app chat. Include the playbook name, the exact step name, and a screenshot of the step's input configuration. Bardeen's team can pull server-side logs for the AI calls associated with your account to see whether requests are reaching their backend.

If Bardeen's AI step reliability is a dealbreaker for your use case, consider using Bardeen solely for the scraping part of the workflow and sending the scraped data to a more stable automation platform for AI processing. This hybrid approach is increasingly common among power users.