01What AI Agencies Getting Crushed Means

AI agencies getting crushed refers to the pressure traditional SEO agencies face as clients see AI tools doing the job cheaper and faster. In this episode, Julian Goldie and his guest Charles talk about how agencies must now compete not just with each other but with AI models like Claude and GPT-6 Astra.

A client can ask an AI to write a report or build backlinks and get an answer in minutes. This makes agencies look slow and expensive. The only way to survive is to be far better than AI alone and to set clear expectations with clients.

02Main Forces Hurting Agencies Right Now

  • AI replaces the middleman: Clients ditch $5,000/month agencies and try Claude instead. They grade agency reports through AI, which often finds fake flaws.
  • Tool costs keep rising: Ahrefs and Semrush raise prices and add subscriptions, but free open-source projects like OpenSEO give comparable data for $0.
  • Link network crackdowns: Google targets large networks (like Outrank) and penalizes thousands of sites, including innocent ones that linked to them.
  • Personalization changes search: Google now builds a personal profile from your photos, emails, and browsing history. This changes how results and clicks work.

03How to Set Up a Hybrid Agency Model

Your agency must combine human expertise with AI speed. Here is the step-by-step process discussed in the episode:

  1. Identify your core service: Start with a standard SEO and link building package, then add consulting on top.
  2. Use AI for heavy data work: Tools like Muse (priced at $0.10 in, $0.20 out) can compile 16,000 queries and analyze 100 pages per query. Use them for the first pass.
  3. Keep a human expert in the loop: After AI gathers data, a seasoned SEO (10+ years) makes judgment calls on link quality, page authority, and strategy.
  4. Add a caveat section to reports: At the bottom of every report, explain what the numbers mean. This stops AI from misreading the data and claiming your work failed.
  5. Set expectations upfront: Tie sales commissions to long-term client retention, not one-time deals. This encourages honest selling.

04How to Use AI in Your SEO Workflow

Julian and Charles walk through real examples of how they use AI to save time and improve results.

  • Large data analysis: Charles feeds 16,000+ queries into Muse. The model extracts ranking features from 100 pages per query. Then he runs that data through Fable or Astra for final reports.
  • AI computer use: Astra’s computer use mode can load 108 pages in 20 seconds. Claude used to take 5 minutes. This makes speed a competitive edge.
  • Cross-checking answers: Julian asks the same question to 3 different AIs (Claude, ChatGPT, Astra) to compare results, because no single model is reliable all the time.
  • Vibe coding: After losing his developer team, Charles learned to vibe code himself. Now he builds his own dashboards and tools without relying on expensive platforms like HubSpot.

05Results and Comparisons Seen in the Episode

Charles shares concrete numbers and tests from his work.

  • Muse vs DeepSeek: Muse costs $0.10 per call (input) and $0.20 (output). DeepSeek costs $1 to $2 for the same work. Even if Muse is 50% as good, the token volume is hundreds of percent higher.
  • Astra vs Claude speed: Astra loads 108 pages in 20 seconds. Claude took 5 minutes. Astra is about 4–5 times faster at computer use tasks.
  • Community launch: Charles hit his 30-day revenue goal in just 6 days by building a paid community for AI search solutions. His 50-person survey showed people want centralized, practical advice.
  • Outrank network size: MyBlogGuest had 23,000 sites penalized. Outrank is bigger — potentially tens or hundreds of thousands of sites. Collateral damage already happened to sites that never used Outrank but linked to users.

06Tips, Limits and Mistakes to Avoid

Charles and Julian share hard-learned lessons from running agencies and surviving two Google-wide bans.

  • Don’t be too loud with link tactics: Google targets the biggest promoters of a tactic, not the tactic itself. Being famous for a PBN or network makes you a target.
  • AI grading is flawed: If you tell an AI to find holes in a report, it will find fake issues based on old SEO advice from 2015. Always include a caveat section in reports to counter this.
  • Salespeople must be aligned: One-time commission on a sale pushes them to oversell. Tie commissions to long-term client retention instead.
  • Separate Google accounts: If you run multiple sites, use different Google Analytics and Search Console accounts with separate proxies. A single penalty can wipe your whole portfolio — Charles lost two accounts this way.
  • Old AI training data is a problem: Many AI models still rely on SEO advice from 2015–2016. Their recommendations are often outdated. Always validate with current knowledge.

07Who This Matters For and Why

This episode is for anyone running an SEO agency, freelancing in SEO, or building a business that depends on search traffic.

If you are an agency owner, you must offer a hybrid human-AI service that is 10x better than two years ago. Your deliverables must be faster and more accurate. If you are a solo SEO, combining your years of experience with AI gives you an edge over both pure human teams and pure AI setups.

As Charles says, the person in the middle — using AI but making decisions from 10 to 20 years of expertise — will win.

Clients don’t care about link types or semantic markup. They only care about more leads and more revenue. Tie everything back to that goal, set realistic expectations, and never let AI grade your work unsupervised.