Transcription
Most agencies spend 60 hours or more and thousands of dollars in labor to rank a local business. Even with AI helping write content, you're looking at 30 to 45 minutes per page, times 30 pages, if not more. That's a lot of work per client before the full system is even in place.
Now, I built an AI agent that runs this entire process. We start with entity research, GBP audit, gap analysis, content production, schema, images, video generation, YouTube upload, and WordPress deployment. All of it. About 90 minutes of runtime under a dollar per page in API costs. And it's bring your own key, no markup.
I'm going to walk you through the exact five-step process my agency has used to rank hundreds of local businesses since 2016. And after each step, I'm going to show you this AI agent executing it in real time.
Now, this is the rank map for the test site we used. A GBP for a plumber in Molden, Massachusetts. Average position 6.18. Only 15% of the map in the top three. Look at all that orange sevens, eights, barely cracking the top three anywhere except right on top of their address and a few random spots in the outskirts.
Now, by the end of this video, I'm going to show you what the same rank map looks like after we ran this system. Excuse you. But first, let me show you what this agent actually looks like when it's running. Because once you see it, you'll understand why most of what agencies charge for is about to become irrelevant.
This is what it looks like when you're uh first go to it. So, I'll just blur my email address and enter in here. Okay. And when we first log in, we have a client overview. So, we'll hit the test site here. And this is the main interface. Across the top is all the things that we want to do. We start with settings, categories, crawl, research, and on and on.
So, step one is understanding what Google currently thinks your client's business is. Now, [music] since 2018, Google doesn't match keywords. It matches entities. Your GBP is an entity. Your website is an entity. Every service you list, every category on your GBP is an entity. Your geography is an entity. Google is constantly checking whether all of these entities match. The closer they match, the better you're going to rank. The more gaps between them, and the less Google trusts you.
So, the first thing we do with any new client is entity research. What categories does the GBP have? What's the primary? What are the secondaries? What are the competitors using that this client isn't? Because every missing category is a search where Google is going to show your competitor instead. Here's how this tool works with that. We'll come on out to GBP categories and we'll type in plumber Malden and we'll run the category audit.
Now, what the tool is going to do is literally go and search for the businesses around Molden, Massachusetts. And it's going to say 68 plumber category that 68 times. So that means there are 68 GBPs in Malden, Massachusetts with plumber as their category. Now those plumbers, they used electrician 17 times, heating contractor 14, bathroom remodeler eight, etc., etc. So all we have to do is go down this list and see if any of these categories make sense for this client, for this business, for what they offer.
So now we know what categories we want to associate with this business. But knowing those entities means nothing if the website doesn't back them up. So that's step two, and this is where most local businesses are completely exposed. The GBP might have four categories and 30 services, but the website has a homepage, an about page, a contact page, and maybe a service page that lists a handful of those 30. It ends up being five or six pages trying to support 30 or 35 entities.
Google sees this mismatch, and it's not just Google anymore. It's Google's own AI. It's ChatGPT, Perplexity. It's Claude. They're all evaluating the same entity signals. When you fix this for Google, you're simultaneously becoming visible to every AI model that's starting to replace traditional search. If your client's website has fewer pages than their GBP has services, then you know what the problem is.
So, this is why at my agency, we developed the core 30. That means one page for every category and service that's on the Google business profile. The homepage targets the primary category plus city. Each service gets its own page and the internal linking follows the same hierarchy that we have on the GBP. Homepage to the secondary categories and the secondary category down to the service pages under that secondary category. You basically build a website that is an exact mirror image of your Google business profile.
But you can't build that structure if you don't know what's missing. So that's where we're going to come in and do a gap analysis. [music] Most agencies skip this entirely and it's a big part of the reason their clients are stuck for months.
So to do the gap analysis, we start by pasting in categories and services. So we're just going to literally copy paste in any format you choose and hit parse and the tool will parse out the categories and services on your GBP. Then we're going to come over to this crawl tab and we'll hit crawl. And once it finishes crawling, uh, it will give you all the pages on the website, including what the AI tool thinks is the correct service or category for that page. So, for example, here we could say that page is for plumber. And if the tool is decided, we're going to say auto on it. So, we can select manually. Uh, we can let the tool decide, but the point is we go through and make sure this is correct. And then we come on over to research. And what this is going to tell us is that we have P pages for all of these services and categories, but we're missing pages for these. So, we want to import those to the bulk content generator, get them on the queue so that we can generate content for them later.
Okay, so that gives us the gap analysis. So, we now know what the core 30 looks like for this client. But for a lot of markets, the core 30 alone isn't enough to dominate. We need to go further. We need to build more relevance. And this is where most agencies have no idea what to do next. So their suggestion is weekly blog posts, which is a waste of time. Do not do weekly blog posts.
There are two additional types of relevance that we need to build. Topical relevance and geographical relevance. Remember Google matches entities. Topical relevance connects your business entity with your service entity. It proves you actually do what you say you do. Geographical relevance connects your business and service entities with a specific location entity. It proves you operate where you say you operate. You need both. Most agencies build neither, but we need to start with topical relevance.
So, we search for the target service and pull people also ask questions. We grab competitor headlines that are already ranking. We find Reddit threads, real people asking real questions about that service in that location. And all of this feeds into additional content topics beyond the core 30 pages that answer the exact questions your client's customers are already typing into Google and typing into ChatGPT. Doing this manually, even with AI helping, can take several hours per client.
Let me show you what the tool does. So, going to come on over to supporting planner, and I'm just going to pull up the past run so that we don't have to sit and watch it run. It takes a few minutes. You'll type in the service term and hit research. And then the tool is going to scrape the people also ask questions. It's going to scrape Reddit and find these. It's going to scrape other local forums and you'll go through this list and think, are these actually relevant to the search term plumber for people who are looking for a plumber? So, how much should a plumber charge per hour in 2026? And if you don't like it exactly, we can click on here and we could maybe change it to how much should a plumber charge per job in 2026 and save. And now that becomes the the term the scoring is based on the AI tool scoring it. How relevant does the tool think that particular query is for the service term? But obviously I probably wouldn't do something like DIY plumbing versus hiring a plumber. Uh, but any ones that you like, we just tick the boxes and we hit send to bulk queue.
Now geographical relevance. I want to show you how we used to do this manually. So the concept is very clear before you see what the tool does. So we're going to start by looking at this local rank map. So, if you haven't seen one of these before, this is telling you if you were standing in the places where the dots are and you searched for window cleaning Las Vegas, this is the position that this particular business would be located in. Right? So, here a lot of one, twos, and threes and then we very quickly fall off.
So, now if I need to build geographical relevance, I need to convince Google that, hey, right where this four is, I need to convince Google that we actually clean windows in that location. So, I'm going to zoom in and I'm going to be looking for landmarks that are on Google Maps. So, this is a Google Maps based interface. This is the tool Lead Snap. And with this, anything that's on this map, we know Google recognizes because this is Google. So, Highland Falls Golf Club, we would write an article that says window cleaning near the Highland Falls Golf Club in Las Vegas. Window cleaning in near Timberline in Las Vegas. Window cleaning near the Police Memorial Park near Las Vegas. We're basically looking for these hyper local signals to convince Google that we do window cleaning in this area. And all of those articles would be based on trying to turn this four into a three because I would much rather turn a four into a three than turn a 10 into a nine. Nine is still losing, not getting any traffic.
Now, let me show you what the tool does. So, I'm going to come on over to Geo Planner. You basically start by uploading your local rank map uh from Lead Snap. And uh, this is designed to work with Lead Snap. If you've not used Lead Snap, um, it's a very good tool. We use it for GBP management. There's a link in the description if you want to try it. You'll get half off your first three months. It's an affiliate link, so if you don't want to use it, no problem, but it will give you that discount. So anyway, you'll upload your local rank map and you'll put in your target service and the city. And then you'll hit generate places. And what the tool is going to do is analyze all of the neighborhoods on it. These directions are how far away and in what direction these places are from your GBP. So, Medford, Mystic Avenue corridor is 3.3 miles east of the GBP address. And here's the reason why the tool thinks that this is a nine for 10. [music] Uh, rank six single well-optimized location page targeting landmarks could push into the top three competitor density blah blah blah. So, it's giving you the reasoning why it wants to target this area, why it wants to target this area. Once we decide whether we agree or not, we scroll down here and here's all the content around that. So, we have the Medford City Hall. We have Hormel Stadium. So, this is a yoga studio. We don't really want to write stuff about the yoga studio. So, we'll untick that. Uh, the just the general Mystic Avenue corridor, the Lego Discovery Center. So, we can go through all of these different landmarks, decide which ones we like, and send to the bulk content generator. And it's important to note that all of these landmarks are being directly pulled from the Google Places API. These these are coming from Google's own database. This is not guesswork.
So now we have every page that needs to exist. We have the core 30, the topical relevance pages, and the geographical relevance pages. But knowing what pages to build is half the problem. Because if the content on those pages reads like everything else on the internet, Google has no reason to rank it. So how do you fill all these pages without sounding like a robot? Two things kill local content. First, it reads like every other plumbing page on the internet. Then Google basically has no reason to rank it. It actually has no reason to even index it. And indexing generic AI content is becoming more and more difficult. The content needs to be distinct. It needs to add something that isn't already out there. Secondly, it needs to be local. Not "we serve the Malden area" local, but actually local. A plumber in Malden should be talking about the triple deckers in Medford, the old pipe infrastructure in Somerville, the coastal weather coming off of River Beach. That's what makes Google and the AI systems believe you're genuinely local, even if you've never set foot in that city before.
So before the agent writes a single word, it does multiple rounds of research. First, it goes out and finds what real people are actually asking about each service. Reddit, local forums, competitor sites, people also ask results. Second, it researches the local area itself, neighborhoods, landmarks, uh, local conditions, anything that makes the content sound like it was written by someone who operates in that market. But research alone isn't going to fix the biggest problem with AI content. And that's because of how AI actually writes. Let me show you what the AI agent research looks like, and then I'm going to break down the writing system that makes this content pass, get indexed, and rank while everything else fails.
So here's the bulk queue and we can see it's in the process of running right now. It's generating some outlines for different articles. And if I just go and open up this one, which has research done, it's going to give me the brief, everything that it found, the PAA, the competitor analysis, featured snippets, local context, all of this information. Uh, it researched for this article before it ever started even writing. Then it wrote an outline before it finally started producing content. And even once all that research is done, writing is where a lot of AI content tools basically fall apart. Uh, you have a prompt, you give it to AI and it spits out an article in one shot. One prompt, one output and it reads like AI wrote it because it did. One model, one pass, one consistent tone the whole way through. That is not how humans write. Humans write section by section. They take breaks, they come back, they revise. The tone shifts slightly between sections. Sentence lengths vary. That's what makes writing feel human. Boy, you look like you are really agreeing with me here, huh? You just keep looking up at me.
Let me show you how this tool writes. The agent uses an eight-pass pipeline for writing. Uh, and that engine is different for each of the four main types of content. So, we have service pages, category pages, location pages, and supporting content. And the writing prompts the system uses are different for each one of those. Let me just show you service pages as an example. So, I'll open it up. So the first thing we have is overall content guardrails. So this is a shorter prompt that is put on top of every other writing prompt to make sure that the uh AI doesn't run off and do something completely different. So this is just reminding it every step of the way what it's writing, why it's writing that.
So we talked about pass one is the research synthesis. All of that topical and local research that we just ran. It's raw Reddit threads, people also ask results, competitor angles, local landmarks. This first pass takes all of that and compresses it into a structured content brief that the AI will then use in all of the future writing. What questions need answering? What local details to weave in? What angles are the competitors covering that this page needs to cover, but better? This is the pass that turns a pile of research into a writing plan.
Then pass two, this is our strategic outline. The system takes that brief and builds the full page architecture. Every H2 heading, the angle for each section, the flow from top to bottom. This isn't just a list of headings. It's mapping out what each section needs to accomplish and how they connect to each other. Most AI tools skip this part entirely and just start writing. That's why their content feels like it wanders.
Pass three is the section draft. And this is where it gets really interesting. The agent is going to write each H2 section with an independent call, a separate prompt for each section with a separate call to the API. That results in a slightly different tone and angle. This is the foundation. It mimics how a real writer would draft an article section by section over the course of a day. You come back and each section feels fresh. Your energy is different. Your word choice shifts. The natural variation is what makes content feel like a person wrote it instead of like a machine wrote it.
Then we're going to go to pass four, burstiness. This is one that most people have never heard of. AI typically writes in a very consistent rhythm. Sentences tend to be about the same length. Paragraphs follow the same cadence. That's not how humans write. Humans write a long sentence, then a short one, then two medium ones, then a fragment. Pass four goes through the entire draft and breaks up that robotic rhythm. It varies sentence lengths, mixes in short, punchy lines, adds the kind of irregular pacing that makes content feel alive instead of generated.
Pass five. Now, we're going to inject perplexity. The system finds predictable AI word patterns and replaces them. If something says "significant improvements," it rewrites that to something a human would actually say. Same meaning, but a completely different word choice. AI has favorite words. Robust, leverage, streamline. Humans don't talk like that. This is the pass that kills that AI sniff test. And yes, it does remove the M dashes.
Now, pass six is what we call human bookends. The system writes the first two and the last two sentences of the article with extremely conversational, opinionated language. Those sentences are the ones that Google's algorithm weighs the heaviest. And all of the other AI systems, they're the ones searchers, in fact, read first. They'll typically read the first couple of sentences, then scroll to the bottom and read the last. Getting those right matters more than the rest of the article combined. So, we have a pass focused specifically on that.
Pass seven is focused on conversion because what's the point of ranking a page if it doesn't make the phone ring. So this pass goes through the content and naturally injects calls to action, [music] phone numbers, and conversational language. Not in a spammy way, not in a banner ad dropped in the middle of a paragraph. And each type of content, this conversion pass is going to look and feel very different because it's customized for that type of content. It's basically doing the kind of lines that make a homeowner stop reading and actually pick up the phone. Most AI content tools can produce informational articles. This pass turns them into pages that will generate leads.
And pass eight is the final check. We've had a lot of different calls writing this article. So pass eight just re-evaluates the entire article to make sure it's correct, it's cohesive, it follows the original guidelines, the original content brief, and the original outline. Did it do everything that it was supposed to do? Is the word count where it should be? Are there any leftover AI patterns that slipped through the earlier passes? Anything that doesn't meet the standard gets flagged and rewritten. This is the quality control layer that catches what the other seven passes may have missed.
And while those eight passes are running, the agent is simultaneously running these parallel steps. We're generating an FAQ section. We're generating the meta title, description, and H1 tag. We're generating schema markup. We're generating images with specific image generation prompts. [music] And we're inserting highly relevant external links to authority sources. All in parallel. So by the time the content is finished, the page is fully built. Then the tool will go in and actually render a video script, generate the video, generates a YouTube title tag and description, uploads the video to YouTube, and publishes the whole thing to WordPress with a video embedded on the content. Fully deployed, no human needed. That's the full pipeline.
But you probably want to see what it actually looks like. So let me show you. Here is what a completed run looks like. We have 29 articles completed, zero failed, 53,000 words of content with an average AI detection score of 39%. We can download all of them as a DOCX. This also will publish to WordPress, but we did this run to DocX. We can download each individual one. Uh, we can download the images. And if we come down, we can actually see what happens. We had a couple of failed videos. Uh, but we had some videos that were ready and fully generated. And we can see the AI scores. So, this means my editor can now come in here and know that this 2,000-word article scored 10%. So, it's probably fine. But this article scored 65%, so it's probably going to need another pass to get that AI score lower. But overall, like none of this content has ever been touched by a human. This is straight out of the tool.
Let me show you what one of the WordPress pages actually looks like that's been finished. So, this is a finished page. Uh, we have the hero image, uh, H1. We come through here. We have a table of contents. Uh, and remember, no human touched this. This was fully generated, built by AI, including putting it on WordPress. That's kind of funny for septic service. And here's the YouTube video. I'll go ahead and mute it, and we'll bump up the playback speed so you can see. Uh, but this is the video that the tool created, uploaded to YouTube, and embedded. Uh, we know that these videos improve indexing on Google. They improve ranking. They improve authority. So, this is definitely something that we do regularly with a lot of our content for clients.
So, remember that rank map I talked about? Here's the after. Average position of 3.22, 74% green coverage across the entire greater Boston area. Malden, Medford, Somerville, Cambridge, Revere, all the way out to Lynn and Saugus. And the spots that aren't green, they're ranked fourth and fifth. One more round of geographical relevance pages and those are going to be green, too. This isn't 6 months of blogging. It's not hundreds [snorts] of backlinks. That's this system executed by this agent in a single session.
And this isn't a demo I put together for this video. This is a real tool. Let me show you the actual usage dashboard. We have over 1,000 pages generated by real users. 114 agency owners or business owners running this tool. So remember, 1162 pages. And if I look at token usage over the lifetime of the tool, we've spent $700. That's well under a dollar per page in agency costs. At my agency, my cost per article is under half what it was before. And the quality has actually improved. And what we see happening, most of our labor is now in double-checking, reviewing the content that the tool is producing instead of actually copy-pasting prompting.
So, if you want access to this tool, every single prompt behind it, and a weekly live Zoom with me where I can help you implement all of this for your clients or for your business, it's all inside my AI SEO Pro community. I'll put a link in the description. There's no extra charge for access to this tool and it's bring your own API key, so I don't make anything on usage. But here's the thing, none of this matters if you can't get clients in the door. I want to make sure that you understand the full A to Z system for how to do local SEO. So, I'm going to link you to this video here, which is over 40 minutes long, and I break down every little thing that my agency does to rank local businesses. The video was produced before this tool, but it helps you understand the foundation of what this tool is doing before you come in and start using it.