AI SEO Audits Help Him Manage 3.3X More Agency Clients

search engine optimization or SEO is a massive industry and it's seeing a lot of AI being applied because there's a lot of repeatable tasks a lot of human bias and a lot of research that needs to be done these are the kinds of repeatable tasks that the AI of today is particularly good at hey I'm Max the original flow gramar and this is the studio update the show where I tell the stories of flow grammas from across Ned's Global Community for most SEO Consultants a client engagement starts with an SEO audit there's various technical checks that need to be done on every single page of the client's site then there's numerous layers of research that they have to do as well Marketplace analysis competitor analysis content themes and then from there they have to synthesize all that research and turn it into recommendations for messaging on the website keywords for search landing pages and blog pages to attract all the inbound interest all the things that people are actively searching for today so that their page ranks for those Search terms historically that is a mountain of work for an SEO consultant I met with sherik who's been in the SEO game for over 6 years and he showed me an AI use case that automates the majority of this process while keeping him in control so he can leverage his years of experience while doing it all faster to put it in context when he's working manually in his prime on a best month he can maintain 15 clients at a time with the solution I'm about to show you he now has over 50 clients at a time and he's confident he can push that number he uses NN to power the back end of his solution and then air table for his front end now I think the really Pro thing that he did that makes his solution effective is as a domain expert he modeled I almost slipped down steps guys all right sorry L not worth it the neat thing is that each step of this process is modeled on the human process given he's an SEO consultant he's an expert at that process I think that's the first key to success and then the second one is that each step runs and then adds that to a view and air table with a human the SEO consultant can check that easily tweak it and then running the next step so far anyone that knows anything about AI says inputs outputs so logically good inputs good outputs and if you have a multi-step process where you need inputs at each step adding human in the loop at each step especially where there's a domain expert at the wheels leads to good results as you'll see in this interview I think it's also a great example of chat is not the only interface for AI and very often it's not the best interface I definitely think that's a trend we'll see in 2025 and Beyond and without further Ado let's check out sher's use

case hey sherik would you mind introducing yourself as shik I've been running SEO Agency for quite a while and right now I'm more or focused towards building some AI agents specifically for SEO space how long have you been in the SEO space I have been into this SEO space for a long time somewhere around like 5 to 7 years and in that entire process I have been running this entire SEO campaigns for the biggest of the biggest BRS out there and I'm guessing there's a lot of AI kind of working its way into SEO work I would imagine yeah 100% there's a lot of things which is actually going on right now in this Ai and SEO space this AI space it's going a wild out there it's pretty insane the other day you wake up and it's like whole different Shenanigans which goes out there yeah I think Shenanigans is the perfect way to put it I think it's a perfect segue sherik would you mind letting me know what your use case is and what kind of motivated you to automate I was been into this SEO space for quite a while been running this SEO campaigns for me the biggest bottleneck in that entire process was putting together a better research strategy and a te AIT part which usually take me a lot of time even though I was in my prime it was very difficult for me to manage most of my clients and it was very difficult for me to scale then I actually take a took a step back and understood that how would I automate that entire process now since we have done that I can easily manage 50 clients at once without compromising all the quality and we are seeing some insane results with most of our clients you do these reports for these clients as part of SEO work how many were you doing before you Prime at your maximum without AI oh man it was 10 to 15 clients is that a maximum so that was maximum 15 clients that you could support yeah absolutely that was the max I could ever go and that's doing it manually and so now you basically with AI automated the a lot of this report stuff and you're saying you can now do like 50 clients at a time right now there's no number to it I'm pretty sure I can go beyond that as well billables billables I love it sh all right show me the money could you share your screen and show me your workflow I would love to do that for you so what we do here is we first

go ahead and click in your project name specify your url specify your location code for you helping the system understand what is the exact region you want to Target the regionalization sorry sherik for us who aren't SEO experts is a location code does like every country have a location in SEO as a location code no so you can see all of the different locations have different location codes got you okay so this focuses like the research and everything the AI is going to do for that specific region yeah exactly just like the location sheet I've shared you we also have what is the language you want to go and Target on you have that flexibility to go ahead and run an SEO campaign as a poly lingual way you come here in the action part the first step is you go ahead and click on audit Tech issues the moment you hit this button another workflow kicks in which does a URL level inspection of your website you see here at this URL level inspection tries to identify all of the different Trend problems which is there on your website and this problems have been segmented on the basis of on page report be it meta title dead length meta description H1 tags and image all tags technical issues like 3xx issues 4xx issues and everything I'm not going to bore you with all of the different technical chons out here it sounds like I've seen SAS apps that sell just this is that right this is like a whole industry in and of itself of just doing this these technical audits you're absolutely right we are combining screening frog and ah's audit at one you don't have to like juggle between the tools back and forth in order to get your audit report then moreover it helps you identify the PA speed few of the recommendations then you move on to the next part identifying your target audience and really digging deep onto them now coming back to the action tab the moment you hit generate project details another AI agent kicks in order to identify who exactly is your target audience what is your offer and moreover what is the content pillars that needs to be written in order to identify the target audience and moreover once that is done tries to understand how are you different than most of your competitors think of it like an USB of sort to the next action step which is the keyword research you hit keyword research tab now once you do that another AI agents start to do a keyword research for you this keyword research has been broken down into multiple steps first part is the P topic now the moment you hit that button out there the keyword research part the a agent goes ahead and tries to understand what are the different parent topics ideally what you can write in order to attract that set of target audience which identified in the very first step where are these parent topics coming from is it coming from a previous step or is this sort of research that it's doing in the nadn work for it's more of the research phase that's actually happening on the nadn and from there the entire data has been contextualized and been put into the a table part and once we have identified what are the parent topics what are the secondary keywords the competition level the search volume then we try to do something called as segregation of the context what do I mean by that the moment you hit generate topic this is where the actual magic starts to happen this system goes ahead and tries to segment that topic into multiple silos you see SEO application for finance and specific sectors was one of the key topic which we identified in the previous stage the system suggests this very respective parent topic have 30 silos into it and the each silos should have their child topics into it and The Silo topics are tell segmented in a way that you kind of have an understanding of what are the lent permutations and combinations that needs to be written in order to have a topical Authority in the niche and once you identify that okay this is The Silo topic I want to go ahead and get started to create an authority on the next part which is the content creation part the child topic is creating the financial content what are the semantically related keywords it actually went ahead and understood what are the semantically what are the secondary keywords for the fact once that is done and you have identify that okay this is the content I want to go ahead and create a web page on you click on generate blog now the moment go ahead and click on generate blog a research agent has been kicked in order to formulate a unique ranking hypothesis which then will be used by the content writer agent this is the main topic and these are the sematic related keywords if the keywords are not actually correct it will go ahead and remove the keywords from the secondary keywords and it will go ahead and add suggested keywords it checks the coherence whether you have a very specific content with a tight cluster once we are done with that the first process of checking the coherence between the primary ke keywords and the secondary keywords it D deeps onto what are the subs saying for it what are the top competitor links out there on the very respective subs and once the top identification of competitors has been done it moves on to the search intent analysis the search intent for secondary keyword varies so take deeps onto what are the sub intents and the Sub intents inside the main keyword which would definitely be missed by most of the SEO experts out there then once that process is done it tries to identify what are the key subtopics that has been discussed in the top ranking articles on the serbs and once that is done it Dees onto understanding what is that most of the competitors are talking on the subs and what are the things what most of them have missed on the very respective keyword and y yada and then this tool This research agent has also access to perplexity and all the other AI tools whenever it needs to do a more in-depth research it can call those tools by itself and do a research for a better inside for the condent This research this is a step that a human would do when they're doing this report is there right 100% before this exact process the search brief was something which used to take me a lot of time and I bet most of the seos out there struggles in order to go ahead and do this very detailed research this is a raw input that research document is that better than a junior would do of course yes man there's a lot more better than mid-level to advanced level SEO expertise out there and then once we are done with that it highlights what are the outline suggestions and everything how the content should be actually structured once this process is done it take deeps on do finding the unique ranking hypothesis and once this entire information has been collected then passes down this entire research brief to the content writing agent this content writing agent has been fine tuned on your choice of words your branding style I love how this is built in Ed table and thank you so much for not making a chat AI agent this is not a chat use case my first thought was like why doesn't he have this just automatically every step run and I think the answer is going to be because the way you have an air table there's this beautiful like perfect in the loop you as the expert look at this you change a word here you remove that citation you tweak something and then you go to the next step so each step you're getting a human audit a human opportunity to say hey wait no this one's more important tweak this a little bit a couple percent of that human context I'm guessing the results are way better right yeah amazingly better we have been doing this for way number of clients of ours and the results are like pretty insane the articles are like quickly ranked and these guys are pulling some insane torrential downpour of traffic towards their website that's so cool to hear once the entire content is being produced it generates The Meta title meta description the slug and also generates an image prompt and the moment you hit this button generate image it will go ahead and generate a featured image which is again fine tuned on your choice on your branding guidelines and then directly push on your website wow that's super powerful this is the end to end I'm so curious what happens

under the hood I can walk you through the process this is how exactly the system looks like it's like the Behemoth of the automations look at that flow gang so actually the back end is set up from here on the a table automations you can see you're thinking all of the webbook requests and everything you're using air tables basically triggering and to a generic web hook into n and then you got that switch node that's like routing to all the different cases that it could be right yeah exactly now we would love to walk you through the research team and the content writer agent the moment you hit that button it gets a record the edable gets a the card and passes down to the research team and this research team has access to multiple tools and an llm model like scrape side tool the sub scraper and perplexity if I open up this for you real quick this is a research agent and this research agent can go ahead and call the tools multiple times wherever it needs to in order to do the in-depth level of research this is attached to Gemini chat model because we have experimented with open AI as other models as well but we have seen that Google gini Chat model is something which is very good at calling tools now once the entire research has been completed it updates a record into aable and then passes that research to the writer agent now this writer agent has access to again perplexity tool as well let's just say that there is some specific type of resarch that needs to be done before the writing process this writer agent also kind of like pulls the data pulls the research if that's something the writer agent wants to highlight further and then once this entire process is completed it extracts that information and then updates the record for it Sher would you mind if we take a quick peek at the system prompt for the writer agent this is exactly how the writer agent has been positioned we got an example some shop prompting in there okay got you yeah the sources and everything the topic the parent topic the semantically related keywords we pulling all of the different Dynamic Fields out there from here so it has the context on what exactly the content needs to be written or be or what is the entire research and that has been passed through in reference to the keyword they you're basically piping in the stuff from the last air table step then into that prompt so it combines with the static makes a lot of sense this is a very basic system what we have with for most of the B2B and the service based businesses you know word actually at the strategy level SEO changes what do I mean by that when you doing an SEO for e-commerce or for SAS or for an Enterprise of a local business the entire strategy changes so this is just the starting layer we kind of like scratching right now we are like I said we are seeing some insam results out of it I think for a lot of people watching they'll be like wow I didn't realize you could do that with n ATN and with Automation and AI but I totally agree and I think it's a great reason why some of the stuff you're doing here probably some off-the-shelf tools can do parts of it but now you took the time to build out a custom system obviously there's a little bit more overhead perhaps but you can make it perfect for Ecom Medtech you can make it perfect for that hot new industry that comes out a year from now that we don't know about or you could white label this maybe there's some big SEO agency that once they license it has to run on their own servers I'm sure for the right price you could do that as well 100% Max there's one thing there's no limitation on what we can actually do over the top of the system if you want to have a very detailed report which you want to send it over to your client definitely you can do that too there's one more thing I would like to highlight here most of the people out there they see automation as the big thing but I believe automation is as good as your own internal processes if you have a very strong Foundation of what is exactly you want to automate you can't automate and get results what you desire it's a classic example of garbage in and garbage out absolutely and I think garbage in garbage out was one of the things when I was looking at what you were doing some people like when chat first came out people tried right an SEO report obviously was garbage no context and I think what show today is a great example of add context then add context then add then do and how you modeled it and I think this is the biggest advice I can give folks is like structure it into discrete steps like you had a researcher and then a writer I'm guessing that was inspired by the fact that's like the humans that do it in the process already even if it's the same person it's the Hat you put on it's like all right I got my research hat on now my advice to folks is study the human process know it if you're not the one doing it you got to go sit with the person study it helps if you're the person and model your AI agents steps as those people like it's literally a team of Agents getting the work done that's absolutely correct yeah that's what exactly what I've been seeing and there one more thing I would like to highlight is most of the people out there they try to complicate the process they try to have a multi set layer of agent and try to have like a 50 or 60 different agents which is kind of like doing a simple thing but that's not how you should optimize the system they say that the biggest complexity is in Simplicity and I think you did a great level there and I know we only saw part of the workflow because again like I don't know if we can do a 5 hour video all the details you broken down into two steps and I think that's a perfect decision right because you have research and then synthesize it and I think the nice Pro tip gang is when you break down your steps like that as well you happen to use Gemini for both but there could be scenarios where one model is better at that task so maybe Gemini is amazing at tool calling but maybe if it's really important maybe anthropic will come in for writing content sometimes or whatever these models get evolved you can swap them out now Sher since you're bullish on Google Gemini which model specifically did you use for this use case and which recommendations do you have for folks in the Gemini family it's more of towards the 1206 model with a million context I'm not just bullish on Google Gemini I'm more keen towards what is the exact specific use case we're trying to get at let's just say that we want to generate content in a European language what I have seen is mril is very good on crapping those European contextual terms all right and wherein Gemini is very good at calling the tools and doing the research part and then we use Mile in order to go ahead and generate the entire content in like French or inment got you so right tool for the job basically just like in any kind of engineering so sherik thank you so much for your time pleasure is all mine if folks like what they see where could they follow along for the journey I'm going ahead and posting a lot of content on LinkedIn you folks can go and see out there you can reach out to my email which is sharik cab. and we can take things forward from there beautiful well sherik thanks again for your time appreciate it have a beautiful rest of your week mate yeah all right ma' thanks for having me

again how big was that workflow right I think it's very funny in the Inn Community you sometimes say users saying oh I have a problem with my big workflow and the 10 nodes and as an Australian I always think that's not a workflow this is a workflow so if you like what you saw drop a comment and mash that like button really helps make sure I keep my job and Americans will know what I mean when I say this but I went to um Mish so I def got them student loans to be paying off on a Euro salary on that note I'm in a beautiful location in Berlin if you find yourself in htin park in the summer check the flat term it might be a sneaky little Rave happening there if you're lucky but since I was filming up here I'm technically on the cockal don't forget it brother and I almost slipped and fell like four times look at all that ice I'll be checking the workers compuls when I get home with this or other chaty BT will on that note I'm Max this is the studio you're a and happy flow gramming