I Let AI Agents Run My SEO. Here’s what happened… — Transcript
Full transcript
- 0:00Earlier this year I added $19,000 in
- 0:02monthly revenue [music] to a client's
- 0:04site just by getting a crack team of AI
- 0:06agents to find and fix what [music] was
- 0:08broken. Here's what happened. For years
- 0:10I'd been convinced that Google Search
- 0:11Console had been hiding things for me.
- 0:13I'd stare at line after line of data
- 0:15convinced there's untapped revenue in
- 0:16there just waiting to be found. So when
- 0:18the perfect client came along I saw my
- 0:20chance to finally prove it. And I
- 0:22decided to try something a little spicy.
- 0:24I built three AI agents all with
- 0:26different missions and then let them
- 0:27loose on the site. One hunted pages that
- 0:30ranked but didn't get clicks. One caught
- 0:32content that was dying before analytics
- 0:33even noticed. And one found the weakest
- 0:36pages and mapped exactly what they were
- 0:38missing. What the three of them found
- 0:39for my client was absolutely wild. And
- 0:41implementing their recommendations added
- 0:43$19,200
- 0:44in monthly revenue. But that isn't even
- 0:46the most interesting part because while
- 0:48they were digging my agents uncovered an
- 0:50invisible traffic killer where your
- 0:51rankings hold, your analytics look fine,
- 0:53but your clicks just disappear. I'll
- 0:55show you exactly how to catch it. So in
- 0:57this video you're going to learn how to
- 0:59build an agent that finds every page on
- 1:01your site quietly losing you money, arm
- 1:03your agents with prompts that diagnose
- 1:05like a 15-year SEO vet instead of a
- 1:07yes-man chatbot, and build one that
- 1:09catches the invisible traffic killer
- 1:11your analytics will never show you. My
- 1:13name is Matt Diggity and I run The
- 1:14Search Initiative, an international SEO
- 1:16agency managing organic and AI search
- 1:18strategy and growth for some of the
- 1:20biggest brands on the internet. Most
- 1:22people still use AI like a chatbot.
- 1:24They're still doing all the serious
- 1:25thinking and removing all the em dashes.
- 1:27>> How dare you?
- 1:29>> An agent is different. You give it a
- 1:31goal, you set the rules, and it does the
- 1:33work while you go off and take a
- 1:35So when you point an agent at SEO where
- 1:37there's more data than any human can
- 1:39process, the opportunity becomes what we
- 1:41in the industry call huge. Now
- 1:43whether you're a one-person operation or
- 1:45a 50-person marketing department, nobody
- 1:48has the hours including my client.
- 1:49They're a major US-based moving company.
- 1:52They were publishing new content every
- 1:53single day doing everything right, but
- 1:56something was off. Leads had dried up.
- 1:58In the moving business, no leads means
- 1:59no trucks, no trucks, no revenue. They
- 2:01were pulling their hair out screaming,
- 2:03"What the hell do we do?" Well, that's
- 2:04exactly what these three agents work
- 2:06out. Each one looks for a different way
- 2:07your existing content is leaving money
- 2:09on the table. And as you'll see, each
- 2:11one finds something the others can't.
- 2:12More on that later. Now, real quick,
- 2:14before I get into it, my SEO agency, The
- 2:16Search Initiative, is offering $1,000
- 2:18off the first month for a handful of new
- 2:20clients in exchange for a testimonial.
- 2:22Once we get you the results that you
- 2:23want to brag about anyways, head over to
- 2:25theseearchinitiative.com, fill in your
- 2:27details, and we'll get back to you. Now,
- 2:28back to the agents. First, the tech
- 2:30stack. Every agent uses the same five
- 2:33tools. make.com runs the workflow. No
- 2:35coding, just modules you can connect
- 2:37together like self-aware Legos. Google
- 2:39Search Console feeds you your data.
- 2:40Apify pulls what your competitors are
- 2:42doing. Claude does the thinking. And
- 2:44everything lands in a Google Doc ready
- 2:46for your content writer. Let's build the
- 2:47first one. Agent number one, the click
- 2:50gap agent. This agent finds pages that
- 2:52rank on Google, but aren't getting the
- 2:53clicks that they should. Say your page
- 2:55sits at positions three to 10 for a
- 2:57keyword with thousands of monthly
- 2:59searches. People are seeing it, they're
- 3:00just not clicking it. Where you're
- 3:02showing up isn't your problem, it's how
- 3:04you're showing up. Your title or your
- 3:06description isn't giving people a reason
- 3:07to choose you. And every skipped click
- 3:10is a lead that should have been yours.
- 3:11Let's put that in numbers. Say a page
- 3:13gets 10,000 impressions a month, and you
- 3:15move the click rate from 2% to 5%.
- 3:17That's 300 extra visitors. If your site
- 3:20converts at 2%, that's six more leads
- 3:22from a change that takes 20 minutes.
- 3:26Step one, in Make, connect Google Search
- 3:28Console and add a filter. You're hunting
- 3:29pages ranked three to 20, over 500
- 3:32monthly impressions, with a click rate
- 3:34below par. Rule of thumb here, under 3%
- 3:36for positions three to five, under 2%
- 3:38for six to 10, and 1 and 1/2% beyond
- 3:41that. Step two, the Apify module. It
- 3:43pulls the top 10 results for each
- 3:45flagged keyword, titles, descriptions,
- 3:47star ratings, FAQs, snippets, etc. Step
- 3:49three, the Claude module, where the
- 3:51prompt does the heavy lifting. Mine
- 3:52tells Claude, "Here's my page, here are
- 3:54my competitors. Diagnose why nobody's
- 3:56clicking, then write me new titles and a
- 3:58better description." The full prompt and
- 4:00my exact make settings are in this
- 4:01YouTube description. Copy them, they're
- 4:03yours. Everything lands at a Google Doc,
- 4:05findings, new titles, priority, expected
- 4:07impact. And here's what that actually
- 4:09looks like from the client run. The
- 4:10flagged page, the click data, the new
- 4:12suggested titles Claude wrote. Here's
- 4:14the problem, here's the fix. That's the
- 4:15first agent. The second one catches
- 4:17something most people never see coming.
- 4:19Agent number two, the decay detector.
- 4:22This agent looks for content that
- 4:23performed once upon a time, but now is
- 4:25quietly dying. Maybe Google's algorithm
- 4:27has just decided it hates you all of a
- 4:28sudden. Pay no attention, I still love
- 4:30you. By the time you notice the drop in
- 4:32your analytics, you've usually already
- 4:33lost ground that takes months to
- 4:35recover. This agent catches it early,
- 4:37and the build structure is nearly
- 4:39identical to the first one. The Search
- 4:40Console module pulls two date ranges
- 4:42this time, the last three months and the
- 4:44three months before that. The filter
- 4:45flags two patterns. One, clicks down
- 4:48more than 20% between those windows, or
- 4:50two, impressions holding steady while
- 4:52the clicks fall. That second one is
- 4:54sneaky. That's the invisible traffic
- 4:56killer I mentioned before. Ahrefs proved
- 4:58it. When an AI overview shows up, the
- 5:00top result's click rate drops by 58%.
- 5:02Your rankings hold, but your clicks
- 5:04quietly vanish. And most people never
- 5:06connect the dots. Then comes the Claude
- 5:08module, and the prompt does the big
- 5:10brain stuff.
- 5:10>> Check out the big brain on Brad.
- 5:14>> Don't ask Claude, "Hey Claude, what's
- 5:16wrong with this page?" That's how you
- 5:17get generic advice and insanely
- 5:18confident hallucinations. Instead, the
- 5:21prompt hands Claude a full evidence
- 5:23file. Traffic trends, search data,
- 5:25ranking history, summary of the page,
- 5:27the competitor pages, and even which
- 5:29SERP features are showing. Then it
- 5:30forces a diagnosis. Claude has to pick
- 5:33one primary cause. Is the content stale,
- 5:35wrong intent, missing coverage, a weak
- 5:37title, stronger competitors, or
- 5:39technical problem? Pick one, Claude. No
- 5:41No, "You're absolutely right,
- 5:43Matt." One diagnosis, a confidence score
- 5:45out of 10, the evidence behind it, and
- 5:47the recommended fixes ranked by
- 5:48priority. The prompts down in the
- 5:50description of this video. Now, the
- 5:51output tells you exactly what to update,
- 5:53what to add, and what to cut out. For my
- 5:54client, this agent flagged 14 pages over
- 5:56sliding. Acting on the briefs took one
- 5:58page from a low position of 17 back up
- 6:01to position three for its target
- 6:02keyword. The page that had been sitting
- 6:04at position four just a year before.
- 6:05Now, the third agent hunts something
- 6:07completely different. Agent number
- 6:09three, the depth scanner. Not pages that
- 6:11are declining, pages that were never
- 6:13strong enough to compete in the first
- 6:14place. The problem is that on a big
- 6:16site, these pages hide. You publish
- 6:18something, it doesn't take off, and you
- 6:19move on. Meanwhile, your competitors
- 6:21publish thorough, detailed version of
- 6:23the same topic and snag the traffic that
- 6:25should have been yours. Here's the
- 6:26build. The data source is Appify again,
- 6:29but pointed at your competitors. For
- 6:30each of your pages, it scrapes what the
- 6:32top ranking pages for that keyword have
- 6:34written and counts their words. It then
- 6:36scans the quality of competitor content
- 6:38in terms of data, tables, FAQs,
- 6:40headings, etc., and compares it to your
- 6:43pages. Then, Claude maps out exactly
- 6:45what's missing. Specific headings to
- 6:46add, angles a competitors covered that
- 6:48you didn't, or elements like comparison
- 6:50tables and FAQ sections that make a page
- 6:52more likely to convert. Again, the whole
- 6:54prompt's in the description. For my
- 6:56client's site, these agents found 28
- 6:58pages in their initial run. Eight click
- 7:00gaps, 14 decaying pages, and six thin
- 7:02content pages. Within 3 weeks of
- 7:04updating them, they were generating
- 7:05$19,200
- 7:07a month in new organic revenue in GA4
- 7:09from pages that were already sitting on
- 7:10the site. If you'd like us to do this
- 7:12for your site, head on over to the
- 7:13searchinitiative.com right now. Until
- 7:15the end of the month, we're taking on a
- 7:17limited number of clients for $1,000 off
- 7:19the first month.
About this transcript
This page contains the full transcript of I Let AI Agents Run My SEO. Here’s what happened… by Matt Diggity, generated from the public captions YouTube serves with the video. The transcript has 1,538 words across 235 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.