← все видео

How I Rank #1 in ChatGPT, Claude & Perplexity (AI SEO Blog Agent)

Max Mitcham · 2026-04-23 · 10м 55с · 258 просмотров · YouTube ↗

Топики: launch-ai-content

Аудио ещё не скачано.

📝 Summary

model=openai/gpt-oss-120b · prompt=summary-v7 · 3 710→1 557 tokens · 2026-05-28 09:02:18

🎯 Главная суть

С помощью связки OpenClaw, Hermes и Obsidian построен полностью автоматизированный «блог‑агент», который генерирует SEO‑оптимизированные статьи, публикует их в Webflow и уже занимает первые позиции в поиске Claude, ChatGPT, Perplexity и Google AI.

Проверка реального ранжирования

Для подтверждения эффективности автор задал в разных системах запрос «Can you suggest any tools for B2B social listening?». В ChatGPT, Claude (в режиме инкогнито) и Perplexity его сайт появился № 1 в результатах. Это показало, что контент действительно попадает в топ‑результаты AI‑поиска, а не просто подстраивается под личную историю браузера.

Ежемесячный SEO‑аудит и база знаний в Obsidian

Раз в месяц запускается агент, использующий RFS (Remote File System) для глубокого аудита позиций: собираются целевые ключевые слова, анализируются конкуренты, формируется детальный отчёт. Всё полученное сохраняется в Obsidian, который выступает в роли «мозга»‑хранилища знаний. Эти данные затем автоматически подхватываются при создании новых статей.

12‑шаговый интервью‑процесс и структура контента

При запросе новой темы агент проходит 12‑шаговое интервью с пользователем, уточняя аудиторию, угол подачи, авторов и т.д. На основе полученных ответов формируется pillar‑статья (длинный, охватывающий тему полностью) и 3‑4 cluster‑статьи (короткие, раскрывающие отдельные аспекты). Такая иерархия усиливает внутреннее перелинкование и повышает авторитет страницы.

Исследовательский агент: Firecrawl, параллельные суб‑агенты и генерация изображений

Для сбора свежей информации агент одновременно запускает три параллельных суб‑агента:

  1. Обращаются к прошлым записям в Obsidian.
  2. С помощью Firecrawl сканируют веб и находят новые релевантные источники.
  3. Генерируют план статьи (структуру pillar и cluster).

Далее пять независимых писателей‑агентов пишут тексты одновременно. Для визуального контента используется ChatGPT Image Model 2, а также скриншоты, полученные Firecrawl, которые автоматически встраиваются в статью. Результат — полностью готовый, визуально обогащённый пост.

Автоматическое создание и публикация в Webflow, кросс‑линкинг

Сгенерированные статьи сразу отправляются в CMS Webflow в статус «Draft». Агент автоматически добавляет кросс‑линки: pillar‑страница ссылается на все cluster‑статьи и наоборот, а также на любые релевантные материалы из базы. После предварительного просмотра пользователь может одобрить публикацию вручную или запросить правки, которые агент вносит и снова отправляет на утверждение.

Навыки (skills) в системе Hermes/OpenClaw и практические SEO‑правила

Все действия управляются набором skills в Hermes/OpenClaw. Основной skill «Blog Creation Pipeline» содержит ссылки на вспомогательные навыки:

Пример практических рекомендаций, которые агент автоматически внедряет в тексты:

Перспективы полной автоматизации и роль человека

Автор планирует открыть доступ к исходному prompt‑у, чтобы любой мог воспроизвести схему на базе Hermes/OpenClaw. При наличии правильно настроенного Obsidian‑мозга система может полностью автоматизировать процесс: от получения SEO‑отчёта до публикации, генерируя 20‑30 статей в месяц без вмешательства. Тем не менее, на текущем этапе сохраняется human‑in‑the‑loop: пользователь проверяет черновики и одобряет их, что позволяет контролировать качество и вносить творческие корректировки.


Эти разделы охватывают весь рабочий процесс «AI SEO Blog Agent», от проверки ранжирования до потенциальной полной автоматизации, предоставляя достаточно деталей для воспроизведения или адаптации системы без просмотра оригинального видео.

📜 Transcript

en · 1 716 слов · 26 сегментов · clean

Показать текст транскрипта
Over the last 60 days I've been working with OpenClaw and Hermes to create a blog agent that is helping me rank live in things like Clawed, AI search within Google, Gemini and Perplexity and so on. And I'm going to show you in this video how I have gone ahead and created this entire agent to help me rank in these LMS. So before we do that let's see if... this is actually the case and if I'm telling the truth so we're gonna ask it a question that I've been targeting pretty hard which is can you suggest any tools any social listening tools that are good for B2B and let's see what that comes back with boom Straight away, we have come up as number one, which is pretty cool. Now, let's go check LinkedIn. Sorry, not LinkedIn. Let's go check Claude here. Now, what I'm going to do is I'm just going to go into incognito mode so that it doesn't use any of my history because obviously if it uses my history, it will just suggest us. So we're going to say same thing. Can you suggest some tools that are good for social listening for B2B companies? Let's see what happens here. Suspense. Oh, looks like we... Boom! Got a little mention there as well. I thought we weren't getting one there. I got a little bit nervous. So we're in here as well. So, you know, I noticed this the other day, right? We started having a lot of demos come to us and we noticed that people were finding us through things like Claude, ChatGPT, perplexity and kind of Google's AI search as well, which is really interesting. And how we've basically been doing it is I built using Hermes and OpenClaw this kind of like blog. agent so here's how the actual flow works so initially i prompt it in discord in fact let's take a step back we initially have an agent that runs once a month that uses rfs and performs a deep audit on how our seo is performing and it produces an unbelievably complex and detailed report giving me the keywords we need to rank for how our competitors are doing so on and so on so that gets ingested into our knowledge base and we're using obsidian as a trigger fly kind of brain and it's unbelievably wild inside of there right now and i'll do a video on that another day but it has that information already done there so when i come to it and say hey listen I want to go ahead and create a blog around XYZ. It first of all goes into a 12-step interview process. Now the reason that it's doing this is it knows based on the skills and we'll go in and have a look at this that it needs to create a pillar blog and then I think it's three or four cluster blogs around that pillar blog. okay so that's very very key we create a pillar one which is typically longer and then you have your more short form short form blogs which are the cluster blocks then we use this research agent so it basically goes and use three parallel agents in tandem which is calling on past knowledge based on our obsidian memory it's using firecrawl to go out and find fresh new information that are interesting and starts to create the research and starts to plan obviously the pillar and the clusters so like what that would actually look like how that would be shaped and how they would basically articulate the argument from there we then start to build the blocks so it writes five parallel writers in tandem. So it basically uses sub-agents to then go ahead and create these five blogs. While it's also doing this it uses the new ChatGPT image model 2 which is insanely good to create visuals on the fly not only that is it's also using screenshots from its research with fire crawl where it can then embed that inside of the blog as well so we're getting this full kind of really nice well articulated article that's highly SEO optimized along with custom images and screenshots where it needs to kind of strengthen a particular argument or whatever From there, it goes straight into Webflow as a... into our CMS, right? And it goes into Draft. Now, here's the key thing is it's cross-linking pillar to clusters and any other blogs that are relevant as well. It then basically says, okay, hey, Max, these have been drafted. Do you want to go and improve them? I essentially will go... will go into our web flow and have a read and approve or not and if I don't I can come back to my agent and say let's go ahead and make these changes and then it gets approved and out it goes so that's the kind of flow in its entirety let's have a look how this kind of plays in action when chatting with my agent okay so we are now live in my discord we can see all of the different agents that i have i have a lot of different ones funny enough you can actually see that i said i'm going to do a youtube video around this can you give me some help with the hook statement what i should follow anything else key and create me a visual as well so you can see that i'm using this in practice so what i'm going to do is i'm going to say i'm going to fire up a chat here and we're going to say i want to write a blog about you know social signals you can pull from hacker news for GTM teams so there we go let's pull up this thread here and we're gonna see the initial phase and if we look at the initial phase that was this kind of interview process so the first thing that it's gone and done is it's gone and looked at the brain Then it's gone and looked at the skill that I've created, which is this long skill on blog creation and blog pipeline. There's also cross references in this skill to other skills that I've created as well. So we have how to write highly SEO optimized blogs. We have... the actual how to use the web flow skill and we have just this this kind of like overarching skill and we have the image creation and what it needs to do we have a lot of different skills embedded inside of this one skill or where it references others as well to help me get to that point so you can see here it's now asking me a lot of different questions about audiences angles authors and it will continue to ask me questions if I explored this i'm gonna actually ask it a another side question here just to give you a good example of how this is working um let's say side question how are we writing good seo optimized blogs give me the five key points from within your skills. Let's see what it says here as well to see how it's basically been helping write our current blogs to get us ranked as well, which I think is pretty cool. Okay, so it's just pulled one from the blog pipeline and the AI SEO here. So first one was lead every H2 with a 40 to 60 word direct answer paragraph. AI search for things like perplexity, chat GPT, Google... Google AI overviews extracts passages, not pages. So if your first 60 words under a heading aren't the answer, you don't get cited. So that's why if you have a look at some of our blogs, they have the TLDR at the top and an FAQ block at the bottom, both written as extractable snippets. We also are very stat dense, cited in line with name sources. That helps query pattern and headings and comparison tables. The architecture is obviously, as I've described, it's around a pillar blog and the clusters with the bridge quotes cross links, which helps for the SEO perspective. And then we have the named author freshness and schema as well. So we have the byline with the credentials when it was last updated. These things that all matter for kind of the SEO, AEO, whatever you want to call it as well. So what I'm going to do. is if you're interested in this and how to operate this i'm going to drop a link which gives you the prompt so that your hermes or your open claw or your claw code can basically go ahead and set this blog flow up you might need to edit it for wordpress you might need to edit it if you're using a different kind of uh website hosting tool that's not webflow but for for this it will then allow your agent to set this up and mirror this exactly the key thing though is you do need to have a obsidian brain system set up for best results um and you know where we can go from here which i think is pretty cool is we could actually start to automate this entirely given we have the initial seo report it knows what words it needs to try and rank for it knows what competitors it needs to try and kind of outdo And therefore, I could, in theory, turn this into more of an automated system, where it goes and creates the blogs, creates the ideas. I don't come to it with the idea. And from there, then basically just you create it and come to me for approval or even auto publish it if you wanted to. And in theory, you could smash out 20, 30 blogs a month if you really wanted to. It's up to you. for now i'm going to keep it manual i really like a human in the loop process and like working with it and but in theory we could fully automate this and turn this into more of a fully fledged cron workflow

⚙️ Pipeline jobs

StageStatusAtt.UpdatedError
download done 1/3 2026-05-28 08:49:57
transcribe done 1/3 2026-05-28 09:02:02
summarize done 1/3 2026-05-28 09:02:18
embed done 1/3 2026-06-30 06:43:56

📄 Описание YouTube

Показать
I built an AI blog agent over 60 days that now ranks Trigify #1 in Claude, and gets us cited in ChatGPT, Perplexity, and Google's AI Overviews for high-intent B2B queries.

In this video I show you exactly how it works — live demo, the skills, the full workflow, and a free Notion doc with the prompt you can paste into your own Hermes, OpenClaw, or Claude Code setup to build the same thing.

Link to prompt + skill - https://www.notion.so/trigifyio/How-I-Rank-in-ChatGPT-Claude-Perplexity-Google-AI-The-Blog-Agent-Playbook-34b606a5520e81789faffb32ecf35ff8?source=copy_link

🧠 GET THE FREE PLAYBOOK (Notion doc + setup prompt)
Comment "send it" below or DM me on LinkedIn: https://www.linkedin.com/in/maxmitcham/

🔗 LINKS
• Trigify: https://trigify.io
• Hermes: [your link]
• OpenClaw: [your link]
• Obsidian: https://obsidian.md
• Firecrawl: https://firecrawl.dev

📚 WHAT YOU'LL LEARN
How to rank in LLM search (ChatGPT, Claude, Perplexity, Google AI Overviews) by building a blog agent that writes pillar + cluster content optimised for how LLMs extract and cite passages — not just for traditional Google ranking.

The agent does the full pipeline:
— Monthly SEO audit (runs on cron, feeds an Obsidian knowledge base)
— 12-step interview before it writes anything
— 3 parallel research subagents (brain query, fresh web, structure plan)
— 5 parallel writing subagents (pillar + clusters in one go)
— Generates infographics with GPT Image 2 and embeds screenshots
— Pushes drafts into Webflow with cross-links
— Pings me for approval before anything goes live

The 5 rules I hard-coded into the skill (and why they work):
1. Lead every H2 with a 40–60 word direct answer (LLMs extract passages, not pages)
2. TL;DR at the top, FAQ at the bottom — both as extractable snippets
3. Stat-dense with inline citations (Princeton GEO research: citations +40%, stats +37%)
4. Pillar + cluster architecture with bridge-quote cross-links
5. Named author, freshness dates, FAQPage JSON-LD schema

⏱️ CHAPTERS
00:00 Proof: ranking #1 in Claude live
01:30 Verifying in Perplexity (incognito)
02:45 Why this matters: demos coming from LLM search
03:30 How the agent works (the 5-stage flow)
04:15 Stage 1: Monthly SEO audit + Obsidian brain
05:00 Stage 2: The 12-step interview
06:00 Stage 3: Parallel research subagents
07:00 Stage 4: Parallel writers + GPT Image 2 visuals
08:15 Stage 5: Webflow draft + cross-linking
09:00 Live demo: prompting the agent in Discord
10:30 Side-bar: the 5 SEO rules the skill enforces
12:00 Where this goes next (full automation)
13:00 Free Notion doc with the setup prompt

💡 WHO THIS IS FOR
• Founders who want to rank in AI search without hiring an SEO team
• Marketers curious about AEO / GEO / LLMO (AI search optimisation)
• Builders using Hermes, OpenClaw, or Claude Code who want a working blog agent pattern
• Anyone who wants their content cited by ChatGPT, Claude, Perplexity, and Google AI Overviews

🛠️ THE STACK
• Hermes / OpenClaw / Claude Code (any work)
• Obsidian (long-term agent memory)
• Firecrawl (research + screenshots)
• GPT Image 2 (infographics)
• Webflow (CMS — swap for WordPress, Ghost, Sanity, etc.)

👋 ABOUT ME
I'm Max Mitcham, Founder & CEO of Trigify — a signal-based GTM platform that monitors Reddit, X, Substack, Hacker News, YouTube, and professional networks for real buying intent.

🔔 SUBSCRIBE for more on AI agents, GTM engineering, and building with Hermes / OpenClaw / Claude Code.

#AIAgents #LLMSEO #AISEO #ContentMarketing #Hermes #OpenClaw #ClaudeCode #AEO #GEO #B2BMarketing