The Research Phase: Where Content Is Actually Won
AI-assisted content creation works best when AI supports research and structure rather than replacing the thinking required to understand a topic. A content strategist with a decade of SEO experience says he can spend 60–70% of production time researching highly technical pieces before serious writing begins. His five-phase workflow therefore starts with research, then moves through outlining, writing, editing and final polish, using AI to accelerate each stage without skipping the work that makes the content useful.
His research checklist starts with the question that matters most: what is the search intent behind the target keyword? Get that wrong, he argues, and nothing else you do will rescue the piece. From there he studies which content types Google rewards with top spots, whether top-ranking posts go deep on one subtopic or cover many shallowly, and what rich media elements the winners deploy. He once did this by typing a keyword into Google and studying the results page by page; today he uses Perplexity to accelerate the same analysis, asking it to characterize a keyword's intent based on the highest-ranking results.
For depth, he runs ChatGPT's deep research function with a custom prompt that pulls from scholarly journals, community sites like Reddit and Quora, and top-ranking pages, then presents everything as a research report. While that report generates (deep research can take around 10 minutes) he stays productive in Surfer, setting up a content editor document with his target keyword, choosing relevant competitors, and reviewing the target word count and the fact-based topics sourced from large language models and top-ranking pages. YouTube, he adds, is an under-mined source of firsthand experience that written content and language models often miss, and YouTube's built-in Gemini features let him summarize long videos instead of watching them in full.
The First Big Mistake: Letting AI Choose Your Angle
Here the process hits its most important guardrail. The presenter's first explicit warning is against the temptation to take a finished research report and throw the whole thing into another AI assistant with instructions to write a blog post. He has tried it, he says, and an AI will almost always go rogue, choosing a content angle that either isn't what you wanted or simply isn't very good. His recommendation: read the research in its entirety.
For readers who want to make dead time productive, he suggests a workaround with the same goal: copy the ChatGPT research into NotebookLM, Google's free research tool, and have it generate an audio podcast version to listen to while commuting or walking. For cornerstone pieces (the content that drives the most business) he goes further still, reaching out to experts for short text-based interviews. The advantage, he says, is perspectives nobody has, locked away inside experts' minds; because it expands scope and timeline, he reserves this high-effort technique for the pieces that matter most.
Outline and Draft: Paint by Numbers, If You Did the Work
The outline phase, the presenter says, is like drawing a blueprint before building a house, and the payoff for good research is that the outline practically flows out of you. Once a tight outline exists, he claims, writing becomes like paint by numbers. Deadlines have tempted him to skip the preliminary steps, and that impatience, he reports, only created more work down the line. Surfer helps here too: its outline tab unfurls competitors' outlines, which he studies rather than copies, looking for gaps in his own research and structural patterns worth adapting. He even checks the tables of contents of books on Amazon, since professional editors organize that material optimally.
When drafting, the AI's role shrinks deliberately. He treats it as a writing assistant working in parallel, not a ghostwriter: asking Claude or ChatGPT whether a list of YouTube KPIs is missing anything, surfacing statistics or expert quotes, or requesting a single succinct definition he may use partially, completely, or not at all. The principle he has figured out is that the more volume you outsource to AI at once, the more your unique voice is erased, one definition preserves your style, a whole section erodes it, and entire blog posts destroy it. For writers who do prefer a full AI first draft, he points to Surfer's Write with AI feature, which studies search results rather than generic training data.
Editing: Read It Out Loud, Preferably While Walking
Editing proceeds in two passes. The first is technical: add any missed keywords where they fit naturally, stay within the target word count, re-check the facts page for uncovered topics, and (on existing sites) build internal links, which the presenter calls a hugely underutilized SEO technique that lets backlink equity and authority flow between pages. Formatting matters too: avoid walls of text with plenty of line breaks, and format lists as actual bulleted or numbered lists.
The second pass is for voice, tone, and flow, and it rests on a piece of advice he calls absolutely overpowered: read your content out loud. He stacks two techniques on top of it, reading on a phone, where most audiences will encounter it, and reading while walking, on the theory that movement stimulates parts of the brain that sitting doesn't. The combination, he says, exposes repetition and filler to be deleted, awkward phrasing that creates friction, and grammatical errors. When self-editing hits its limit short of a target content score, Surfer's optimize feature can locate passages needing more depth or keywords, prompting approval for each change.
The Polish Phase: Winning the Only Question That Matters
The final phase addresses what he calls content packaging, and it begins with a blunt question: in a world where anyone can ask ChatGPT any question at any time, why would anybody read content on the internet? His answer is to offer a better, more informative, and more engaging experience than a chatbot, citing an article from Maze as his model, with charts, well-produced visuals, embedded videos, callouts, and firsthand expert perspectives, all things he says a chatbot cannot provide, at least for now.
The packaging toolkit is deliberately accessible: expert quotes from interviews or researched books, embedded videos (other people's videos work as well as your own, he notes), charts and diagrams that drive home core concepts, and animated GIFs demonstrating processes. None of it requires professional design skills, he points to source sites like Envato or Freepik and free tools like Canva or Figma. Once high-quality assets are embedded, the piece goes to a staging site for a final read in context, then publishes.
The full loop has a measurable payoff, by his account: what once took him three full business days per piece can, with practice and these tools, be done in about half a day, a jump to producing one or two high-quality pieces daily. He also discloses a material caveat to his enthusiasm for Surfer, the tool woven through every phase: he makes videos for the company, though he says he used the product well before joining its YouTube channel.