How to Automate YouTube to SEO Blog Post — Auto-Publish with GPT-4o & Whisper with n8n
You record a YouTube video. It performs. People watch, comment, ask follow-up questions. Then it disappears — buried in the algorithm, invisible to Google, never surfaced again in search. Meanwhile, t
You record a YouTube video. It performs. People watch, comment, ask follow-up questions. Then it disappears — buried in the algorithm, invisible to Google, never surfaced again in search. Meanwhile, the transcript sitting inside that video contains exactly the kind of long-tail keyword content that ranks. You just never extracted it. This workflow fixes that. It takes a YouTube URL and, in under 3 minutes, produces a 1500-word SEO article published directly to WordPress — transcript pulled, AI-rewritten, meta tags set, slug clean, post live.
Why Content Teams Are Leaving Video Value on the Table
Most technical founders and content operations teams treat YouTube and SEO as separate channels. They are not. Every video you publish is a piece of unindexed written content. Google cannot rank what it cannot read, and while YouTube transcripts exist, they are not structured, not keyword-dense, not formatted for scannable reading, and certainly not sitting inside WordPress ready to capture organic traffic.
The manual alternative is brutal: download the transcript, clean it up, feed it to ChatGPT with a prompt, copy the output, paste it into WordPress, add a title, write a meta description, set the slug, choose a category, hit publish. That is a 45-minute task per video, minimum. For a team publishing three videos per week, that is over two hours of pure mechanical work — before any editing happens. Most teams just skip it. The blog never gets updated. The SEO opportunity compounds into neglect.
The real cost is not time. It is the compounding loss of organic traffic you never built. A single well-structured article from a high-performing YouTube video can rank for dozens of long-tail queries and generate leads for 18 months. Multiply that across your video catalog and the gap is significant.
The Architecture: How the n8n Workflow Operates
This workflow runs entirely inside n8n and chains four operations: YouTube transcript extraction via the Whisper API, content structuring and SEO rewriting via GPT-4o, metadata generation, and direct publish to WordPress via the REST API. No manual steps. No copy-paste. No waiting.
The trigger is a simple form or webhook — you paste a YouTube URL, the workflow starts. From there, the pipeline runs autonomously: it downloads the audio track, sends it to OpenAI's Whisper transcription endpoint, receives a clean text transcript, passes that to GPT-4o with a structured prompt that enforces SEO best practices, generates the article, creates an appropriate slug and meta description, then pushes everything to WordPress through the built-in WordPress node. The post goes live in draft or published state depending on your configuration.
The workflow handles the parts that break manually-managed processes: timestamp cleanup in transcripts, section inference from spoken structure, keyword density without stuffing, and consistent formatting. GPT-4o does not just rewrite — it restructures. A 10-minute spoken video rarely maps cleanly to a readable article. The prompt is engineered to infer logical headings, convert conversational phrasing to scannable prose, and preserve the original technical accuracy while making it readable to someone who never watched the video.
Step-by-Step Setup in n8n
Step 1 — Configure the Webhook trigger. In n8n, add a Webhook node. Set the HTTP method to POST and note the production URL. This is the endpoint you will hit with the YouTube URL. If you are using an internal form or Notion button as the trigger, you can also use the Form Trigger node directly instead.
Step 2 — Extract the YouTube audio. Add an HTTP Request node. Use the yt-dlp command via an Execute Command node to pull the audio stream from the YouTube URL passed through the webhook body. Configure the node to extract {{ $json.body.url }} as the video source. Output should be an MP3 or M4A file saved to a temp path accessible by the next step.
Step 3 — Transcribe with OpenAI Whisper. Add an HTTP Request node pointed at https://api.openai.com/v1/audio/transcriptions. Set the method to POST, content type to multipart/form-data, and attach the audio file from the previous step. Set model to whisper-1. Add your OpenAI API key in the Authorization header. The response body will contain the full transcript text in {{ $json.text }}.
Step 4 — Rewrite with GPT-4o. Add an OpenAI node (or HTTP Request to https://api.openai.com/v1/chat/completions). Model: gpt-4o. Pass the transcript text as user content. The system prompt should instruct the model to: produce a 1200–1500 word article, use HTML headings (h2/h3), include the primary keyword in the first 100 words, generate a meta description under 155 characters, and return a clean slug. Structure the response as JSON with keys title, slug, meta_description, content. Use GPT-4o's JSON mode ("response_format": { "type": "json_object" }) to guarantee parseable output every time.
Step 5 — Parse and map the output. Add a Set node to extract values from the JSON response: {{ $json.choices[0].message.content }} parsed into its components. Map title, slug, meta_description, and content to named fields you will reference in the WordPress node.
Step 6 — Publish to WordPress. Add the WordPress node. Authenticate with your site credentials or application password. Set the operation to Create Post. Map the title field, content field, and slug. Set the status to draft for review or publish if you trust the output. Add a Function node before this step to inject the Yoast SEO meta description into the post meta array if you use Yoast — the field key is _yoast_wpseo_metadesc.
Step 7 — Optional: Slack or email notification. Add a Slack node at the end to notify your team with the post URL. Keeps everyone in the loop without anyone needing to monitor n8n directly.
What This Changes Operationally
The obvious gain is time. A workflow that runs in under 3 minutes replaces 45 minutes of manual work per video. But the more important shift is consistency. When repurposing content is this cheap, you actually do it. Teams that manually repurpose content do it when they have time — which means inconsistently or never. When it is automated, it happens every time, for every video, on the day it is published.
For technical founders running lean operations, this means your YouTube activity directly feeds your SEO strategy without requiring a content person in the loop. For ops teams at agencies or SaaS companies, it means your clients get more value from every video they produce without proportional effort from your side.
The downstream SEO effects accumulate fast. Each article targets the same topic as a video that already proved interest with an audience. The keyword research is implicitly validated — people already watched it. You are not writing articles about topics that might perform. You are converting content that demonstrably did.
Common Pitfalls and How to Avoid Them
Whisper accuracy on technical content. Whisper performs well on clean audio but degrades on technical jargon, acronyms, and product names. Always pass a glossary or correction dictionary through a preprocessing step before sending to GPT-4o. A simple Find/Replace node that corrects your 10 most common product name misspellings saves significant cleanup time downstream.
GPT-4o hallucinating details not in the transcript. This happens when the prompt is too open-ended. Constrain the model explicitly: tell it to base the article exclusively on the transcript content provided and not to infer or expand claims beyond what is stated. Use temperature 0.3–0.5 for more controlled, factual output rather than the default 0.7.
Slug conflicts in WordPress. If you publish multiple articles with similar topics, WordPress may auto-append a number to the slug, breaking canonical URL planning. Add a pre-publish check using an HTTP Request node to query the WordPress REST API for existing slugs and handle conflicts before the publish step.
Long videos exceeding Whisper's file size limit. Whisper has a 25MB input cap. Videos over 20 minutes may produce audio files that exceed this. Add a conditional step before the Whisper call that checks file size and splits the audio into segments using ffmpeg via Execute Command if needed. The segments can be transcribed individually and concatenated in a Set node before passing to GPT-4o.
WordPress authentication errors with application passwords. If your WordPress site uses a security plugin that blocks REST API access or restricts application passwords, the publish step will fail silently. Test the WordPress endpoint manually with a simple GET to /wp-json/wp/v2/posts before building the full workflow. Resolve access issues at the WordPress level first — n8n's WordPress node cannot work around server-side blocks.
Skipping meta description mapping. The WordPress node creates posts but does not automatically populate SEO plugin meta fields. If you use Yoast or Rank Math, the meta description GPT-4o generates needs to be injected through the post meta array. Missing this means your articles publish without SEO metadata, defeating part of the purpose. Always verify this step is wired correctly in your workflow before going live.
Ja construimos isso pra voce
Nao comece do zero. O YouTube to SEO Blog Post — Auto-Publish with GPT-4o & Whisper e um workflow n8n pronto para instalar — conecta suas ferramentas em minutos, sem codigo.
Instalar por $59 →
