ChatGPT and Claude are fantastic tools for SEO, you can crunch data, speed up your workflow, organise your tasks, and have a personal assistant that can quickly provide answers and solutions. A lot of people are already using these tools in a very basic way, I’m going to show you how you can maximise their usage and hopefully cut out a lot of middle work.
I use AI every day in my own personal workflow, and I’m going to share my insights on how I use ChatGPT and Claude for my various SEO tasks. First we’ll discuss the best use cases, your work setup, and then we’ll dive into specific use cases and how to run them through AI.
Here’s what we’ll cover:
- Competitor research
- Backlink research / outreach
- Keyword research
- Google Search Console
- On-page SEO audit
- Image generation
- Writing
- Technical SEO
- Content Management System
Best use cases 😃
The best use cases for ChatGPT in SEO are data analysis, keyword / competitor research, technical SEO, site management, and acting as a writing assistant. In my opinion these are where you should use Chat and where it knows how and what to do 90% of the time, meaning less oversight is needed (given you have a general understanding of these tasks).
For writing, I would only use AI as an assistant, that means drafting articles, helping with research, maybe editing, interlinking and final checks. I would advise against using Chat or Claude to write the actual copy — I explain this in depth in the writing section.
Workflow setup 🖥️
Before we jump in, let’s talk quickly about your setup so that you’ve got an organized workspace and your AI assistant has access to everything it needs.
We don’t want to be working from the browser or out of an app. For ChatGPT, set up Codex in your terminal: install it with a command line, then launch it. If you’re using Claude, it’s the same process. I’ve linked the docs there; check them out and install them now. It’s super simple—don’t get overwhelmed by the look of things, it’s not complicated at all.
To launch Claude or Codex from your terminal, all you do is open a window, type “claude” or “codex”, and hit enter. It’ll launch the AI and start a session. Works just like the desktop app: type in your question or task and it’ll respond there—it just takes a little to get used to the window.
Yes — the terminal window can look a little weird to non-developer types. If you prefer to customize it so that it is a different color or has bigger text, check out Ghostty; this is what I’ve been using. Open Claude in your terminal and then ask it to download it for you and customize it the way you like.
What Chat or Claude should have access to:
- SEO tools like Ahrefs or OpenSEO (using MCP)
- Github
- CMS (api or mcp for the content part of the website)
- Full access to all your files (downloads, documents, etc.)
- DataForSEO (AI search)
- ScreamingFrog CLI
- Google Search Console + Analytics
For paid tools like Ahrefs or DataForSEO, there are free versions, so pick and choose what you need; this is just my setup. For GSC and GA, you’re going to have to do a custom setup for Chat to have direct access, otherwise you can easily export the data yourself—it’s just not going to be automated (not a big deal).
To start, create a folder on your desktop and clone your entire site or the content pages that you’ll be working on. For example, I’ve cloned the entire blog and category pages. That way I can write on my local device before touching the CMS or backend of the site, and with an MCP, Claude will upload everything for me so I really don’t have to touch any of the backend.
In your desktop folder you can have your data, your code base, research — everything that would be useful context for your AI assistant. I also keep a HANDOVER.md doc to log everything that’s been worked on, and log the future tasks.
Most of the text files will be in Markdown, so download a Markdown editor. I use VS Code because it has lots of uses and it’s just easiest for me (and free), and I write directly in the app.
Markdown formatting is super easy to learn, and if you don’t want to learn, you can tell Claude how you want it formatted and it’ll take care of everything.
Okay, I’m probably missing something but I can’t remember 😅so let’s just jump into the tips!
1. Competitor research
There are a number of SEO tools that can provide insights into how a competitor’s website is performing. We want to see their traffic, the keywords they are ranking for, their brand visibility, AI mentions, their technical setup — everything under the hood. If we have a clear picture of that, we can reverse-engineer their strategies, learn from their successes and mistakes, and we’ll also have a baseline milestone that we can aim for.
In most cases, a competitor is doing better simply because they have an SEO strategy in place, they are following Google’s best practices, and their site is structured & healthy. Regardless, we still want to have those data points so we can know exactly what needs to happen for us to remain competitive. Lots of sites that I’ve audited, the gap was simply that they just didn’t have pages on their websites that their competitors had… very clear solution.
For competitor research keep the process simple: audit your website, audit their website, then compare. Here’s exactly what you should audit:
- Site health (PageSpeed)
- Site structure / architecture
- Backlink profile
- Keyword positions + all keywords
- Distribution of organic traffic
- Number of web pages / best performing pages
- Content velocity (how frequently they are publishing new pages)
ChatGPT can help with all of this. You might need to export data from some third-party tools and feed it to Chat, but with all this data combined, it’ll be able to point out exactly why your competition is doing better, where their weaknesses are, and how you can use that to your advantage. Save all this research in your project folder, as you’ll be constantly referring to this data.
Let me explain how to get the data for each of these (for those that don’t know).
📄On-page SEO - ChatGPT is a web crawler, so you can simply ask it to assess the on-page SEO of competitors' pages, make notes, and compare it to yours. A more reliable way (how I’d do it) is to download Screaming Frog (it’s free) and run an audit on both your site and the competitor's. Download the relevant reports and have ChatGPT analyze the spreadsheets.
🔨Site structure / architecture - The way that your site hierarchy is organized can impact SEO performance. You don’t need any fancy tools for this; ChatGPT can find the sitemap.xml and analyze that. It can then tell if a site has a flat, deep, or hub-and-spoke (best) hierarchy, and then advise you on how to organize yours into more logical groupings. Make sure you really double-check everything here because you don’t want to mess this up.
❤️🩹Site health - Use the Screaming Frog export for this too, as well as PageSpeed Insights. PageSpeed only takes one URL at a time, so take the competitor’s main and best-performing pages and copy the results across to ChatGPT so it can crunch the numbers. Do the same for your site, too.
🔗Backlink profile - There are a number of free tools you can use to check a competitor's backlink profile. I use Ahrefs, but you can try this tool or even the Ahrefs free checker. In my opinion, it’s not that important, but it’s still useful information when you want to start a backlink campaign or understand why their site has so much authority or AI visibility.
📈Keyword positions + all keywords - This is another task for which you’ll need a third-party tool. OpenSEO.so is a free-to-use tool; I would recommend that for a beginner. Not everything is super accurate, but nothing really is. I use Ahrefs, but most of this can be done with other cheaper tools. Export the keywords that your competition is ranking for, the search volume, the Keyword Difficulty, and the corresponding pages. I tend to download all the keywords just to be sure.
📊Distribution of organic traffic - This falls under the same task as above. This is not the keyword volume, but the traffic estimate to each of their pages—specifically, what’s the organic traffic to their homepage, category pages, and blog posts. OpenSEO has estimates, but they are not accurate; from a couple of tests, they are significantly inflated, in some cases by 5x. Ahrefs estimates are a little less inflated, but even still, I only ever treat them as reference points. What to pay attention to: if 90% of a competitor's organic traffic is to their homepage, they are not a serious SEO competitor. Their strength is brand recognition, probably via social media or ads. But if they have a dozen or so pages driving a couple thousand visits per month, take note.
#️⃣Number of web pages / best performing pages - We can easily see the number of pages via the sitemap (ask ChatGPT). If they have a deep site with hundreds of pages all supporting their main service/product pages, that’s a strong competitor. More pages doesn’t necessarily mean better, but it’s a signal to Google, and there are sure to be gaps you’ve probably overlooked. Take note of the best performing pages; I like to look at their /blog/ separately and see which of those are driving the most traffic, as that’s a good place to gather ideas for when you publish on your own blog.
💨Content velocity - How many new pages are they publishing each month? That’s going to give you an idea of their SEO strategy and how seriously they are taking it. A site that’s publishing 20–50 blog posts likely has a dedicated person managing that, and it’s probably going to be hard to match that energy. It also means they understand SEO quite well and mean to dominate. Take note of what they are publishing, too: is it programmatic pages and/or blogs? And take note of the quality: if they are publishing AI trash, it’s actually a good thing for you because they’ll get penalized eventually, leaving an opportunity for a human-written piece to rank.
🤖AI content - This is a new indicator I started looking at. ChatGPT can give estimates on what’s AI-generated and what’s not; it can tell pretty well. You could even use GPTZero, or run a snapshot of their top pages or blogs through ChatGPT to see if their content is heavily AI-driven. If they are using AI to generate most of their content, your competitive advantage will be human-written content, which Google prefers and which will do much better in the long run.
2. Backlink research / outreach
I mentioned how to export the backlink profile from a competitor using third-party tools above; this is honestly the best place to start when planning a backlink campaign. Export the backlink profile for all of your top competitors. Ignore the spam links and pay attention to the blogs and articles linking to their site. Get ChatGPT to compile a list of the blogs or industry publications that link to your competition. If a website publishes a listicle mentioning your competitor’s service, there’s a high likelihood that they’d also add your site there (given they are not a competitor).
Discovering publications within your niche means you know where to pitch your story and also who’s writing about it. You want to build up a relationship with these contacts. ChatGPT is good at searching the web for contacts and similar websites, so you can feed it the info it needs and it’ll go searching and build a spreadsheet of potential contacts that you can email and pitch your stories to. Don’t be surprised if some of these sites ask for payment, that’s pretty normal.
For outreach, when you are ready to send out some emails and pitch your service or product, ChatGPT can be helpful as well. Link it up to your Gmail or email provider and it can draft the emails for you, which is particularly helpful when sending out more than a dozen. Make sure you edit the emails, and make sure you really discuss the angle and approach before sending them out. Anything that reads like AI is going to be ignored. I’ve also found that ChatGPT doesn’t always know the best angle and will push what it finds interesting, but not what a publication or author would find interesting. So make sure you push back and really dig into what’s interesting for their audience.
The most successful backlink campaigns occur when you are pitching something totally unique—not just your app or service, but a report or study based on proprietary data: something that’s never been published before. That’s what’s going to entice a journalist or writer; don’t hit them with the same pitch as your competitor. If you don’t have any unique data to offer, look at the wider industry and see what’s missing. Sometimes it’s as simple as collecting industry data that no one else thought to gather and publishing a report on it.
3. Keyword research
ChatGPT can be a helpful assistant for keyword research because it can do all the admin work, which is tedious and time-consuming. With an MCP integration, Chat can also search and pull keyword lists completely independently, which is actually awesome.
But let’s pull back for a second: if you don’t know how to conduct keyword research, or if you don’t know what to look for or how to identify keyword gaps, then go slow, take your time, and analyze the data yourself. There’s a tendency to offload everything to AI, but if you are totally relying on Chat to do your keyword research, you’re going to miss important details and make some blunders. In SEO, the details matter; a small deviation or a minor detail overlooked can cost you hours correcting the mistake. I’d say this: if this is your business or website, you should be familiar with the keywords and search volumes – because this data flows into Google Ads, user behavior, competitor research, site development, site expansion, and so on.
Okay now let me explain how you can conduct keyword research with ChatGPT or Claude:
- Ask Chat to create a desktop folder for the research
- Give Chat your URL and give it your competitors
- In detail, describe your service or product, what you are selling, which market, all the various services, who the customers are, etc.
- Give all your web data to Chat; this includes Ahrefs keyword lists, GSC data (export), and GA4 (export)
- Tell ChatGPT to save all the raw files you provide it
- Tell Chat that first you want to focus on finding all the keywords that relate directly to the product or service (a.k.a. money keywords)
- Make sure you tell Chat that you want to find the absolute ceiling
- Ask it to provide you with a list of keywords that you can manually paste into Ahrefs (or whatever tool), and export that list + matching keywords + related keywords + different markets for it to review
- Tell Chat to save the relevant keywords to a master spreadsheet (separate from raw files) as you go along, and make sure the sheet includes KD, Volume, CPC, search intent, and market for each keyword
- Chat will crunch these exports and based on that data, it should come up with more keyword ideas
- Continue until you are satisfied that you’ve exhausted the research. Keep prompting chat with different angles.
- Download the keyword lists from your competitors and give that to Chat as well
- Once you’re satisfied that all the primary keywords are documented, you can fan out to sub-topics
- Drill down into each sub-topic that relates to your website and download those lists
- To finish off, I like to build a list of informational keywords for blogs (you have to get a little more creative here)
- Tell Chat to create a separate tab on the master sheet for blog keywords
- Again, drill into each topic and make sure these keywords are adjacent to your service/product; don’t go searching for high-volume keywords that are loosely related, as Google doesn’t like that and it’s not conducive to customer acquisition.
There are MCPs for ChatGPT and Claude that can speed this process up significantly. So instead of exporting lists yourself, Chat can pull them directly. This is great, but you’ll be limited by the number of API calls. Exhaustive research will easily hit your monthly limit. Ahrefs, SEMrush, and OpenSEO all have this feature. If you’re trying to save cash or limit your API calls, do it manually.
4. Google Search Console
Hopefully you’ve got Google Search Console (GSC) set up for your website; if not, go do it now—it takes five minutes. GSC is going to tell you how people are finding your website, including the exact terms that people type and which pages they land on.
You can export all the search data from your site and feed that to ChatGPT to analyze. It can tell you which pages are doing well and which need improvement. It can also surface a bunch of terms you might not see in other tools, but that Google is recommending you for.
GSC will also alert you to indexing issues, which is a major reason why many pages don’t rank at all. Again, you can export the indexing data and feed it to ChatGPT to walk you through the issues and identify what’s going wrong. Make sure ChatGPT cross-checks this against your sitemap as well so that it gets the full picture.
If you want to set up a connector or tool that ChatGPT and Claude can use to pull data directly, it’s a bit of a process, so get Chat or Claude to walk you through it step by step.
Build me a local SEO data toolkit at ~/seo-tools/ so you can pull Google Search Console, Google Analytics 4 and Bing Webmaster Tools data for every property my accounts can see, without me exporting anything by hand. Walk me through the parts only I can do (Google Cloud console clicks, Bing API key), and do everything else yourself. Then eight numbered requirements covering: the venv and gitignore; OAuth as their own Google account with read-only scopes and a reusable auth.py; the exact flags for gsc.py, ga4.py and bing.py; sites.yaml short names filled from --list; JSON caching; a README; a memory note telling Claude to pull these before spending paid API credits; and a proof-of-life pull from each tool at the end.
First, you set up an account on Google Cloud, create a project, and add the APIs for Google Analytics and Google Search Console. Set up authentication, then publish and share the key with Chat. Chat will need to build some Python scripts to pull the data, which will live on your computer. After that, you can query anything related to GSC or GA and your assistant will tell you—it’s pretty awesome. I went one step further and set up the Bing API as well, just to get a full picture.
5. On-page SEO audit
We talked about on-page SEO earlier, and the same process applies here. You can either export data from third-party tools (like Screaming Frog) and feed that to ChatGPT, or simply ask Chat to analyze your on-page SEO. It’s best to measure it against a competitor that’s doing well.
Hey I have this page [URL] that targets [keyword] and here’s the competitor’s page [URL] targeting the same keyword, what are they doing that I’m not doing?
Google’s best practices are widely known, and Chat is totally capable of pointing out why your page is doing worse and suggesting ways to improve. Usually, your keywords are not correct or not in the right places, your URL slug might be confusing, there might be thin content on your pages, or your title tags aren’t correct. Chat will identify these immediately and suggest the right course of action.
I will say ChatGPT and Claude tend to overcorrect: they’ll give you all the things that you could do, which are not always necessary. Putting the keyword in necessary places is important and a must-do, but adding thousands of words at the bottom of your category pages… that’s not always great. Adding an FAQ section to every service page… yeah, if it’s relevant, but not something that’s generated with AI. I’ve even noticed with blogs it’ll suggest adding FAQs at the bottom when it's not important, especially when your blog is already 2,000 words long.
That being said, you should add content to your pages. But make sure it’s relevant, human-written, and helpful. A thousand words of generic copy is a bad user experience and won’t get you far in the long run.
Also, if you’ve identified a whole collection of pages with bad on-page SEO, make sure you go one by one rather than bulk-updating with Chat. When generating a hundred new meta titles, there’s going to be some drifting and missed context, so you’ll likely have to come back and edit them in the future, which is more time-consuming than just doing it right from the start.
6. Image generation
Google loves to see custom images and graphics on pages; it’s a signal that the page is rich and that the author put time and effort into it. It’s also great for Google Images, another traffic funnel you can tap into. ChatGPT is getting better and better at generating graphics and lifelike images. Take advantage of that, and try to avoid the obvious pitfalls of AI images — for lifestyle pics or random graphics, for the most part, it’s as good as a stock library or Figma.
For example, if you were to compare a tutorial blog with text against one that has video content, images, and infographics alongside text, Google is likely to prefer the latter. It’s also nice for user experience. Some readers want visual elements and an interactive page, which usually leads to longer time on page — another signal to Google.
For a beginner, you don’t need to stuff a dozen images into a blog or page. A nice thumbnail image and 2–3 in the body of the article are great. Try to keep image file sizes under 200 KB.
7. Writing
A big part of SEO is writing. Most web pages have around 500-1000 words and those words can play a big role in where your page ranks on Google. When you’re building dozens or even hundreds of pages, you’re working with tens of thousands of words and it’s a lot to write and manage. ChatGPT can help a lot with producing text and managing the text.
A word of caution: I would advise against publishing pure AI text for a number of reasons. First, generated text is generally lower quality than human-written text; it’s also very generic, saying everything and nothing at the same time. Readers are skeptical of AI-written content and are getting better at spotting “AI slop.” If readers realize that your copy or blogs are AI-generated, that will reflect poorly on your brand and create a bad user experience.
Google penalizes AI content mostly for the reasons listed above; user experience is front of mind for them, and where they can tell content is low quality, they won’t serve it to users. Google has also repeatedly stated that they will penalize AI content, which could affect specific pages or even entire websites.
Claude just announced that they will be adding “invisible watermarks” to generated text so that it can be more easily identified. We can assume that Google and every other major search engine will now look to identify those invisible watermarks and factor that in when deciding how your website ranks.
That’s my opinion, though others say differently; you can make up your own mind. Regardless, it’s clear that users want human-written content (for now, at least) and search engines do as well (look at Google’s best practices). So even if you game the system for a few short months, in the long run, you’ll just create more work for yourself.
Okay, now that that is clear, let’s move on to explain how you can actually incorporate ChatGPT into your writing process in a way that’s conducive to long term growth.
You can use ChatGPT for drafting, research, templating, and quality control. This is particularly helpful if you are designing and writing multiple pages at a time, especially if the pages are quite similar in nature. I tend to write a few pages by hand so Chat knows the format, tone, and exactly how I want things done, and then it can draft several more pages for me to edit. For research, best practice is to have proprietary data or a research library (that you’ve already built) and then use Chat to query the library and fact-check as you write. When you are building a lot of pages, you can quickly lose sight of SEO best practices, missing keyword placements, or overlooking title tags, so Chat can help you here by ensuring that the keyword data aligns with the pages you are creating.
8. Technical SEO
Since ChatGPT and Claude excel at coding and development, they are great for auditing technical SEO. They can quickly root out bugs, redirect loops, orphan pages, or cannibalization. Simply asking Chat to run a technical SEO review of your page or site will surface most issues. It's best to go one by one, though, so nothing is missed.
If you’ve got Ahrefs you can run an audit there and then export the report for Chat to analyse and diagnose. It can then walk you through the steps you need to take to fix these issues.
Audit these areas:
- Sitemap
- PageSpeed
- Schema markup
- Breadcrumbs
- Interlinking
- Indexing (Google Search Console)
- Robots.txt (File added to codebase)
- Redirects / loops (codebase +GSC)
- Broken links
- Hreflang (for multi-language sites)
This is obviously going to work best if Chat or Claude has access to your website code, either in GitHub or locally. It can make the updates for you without a developer or technical person stepping in. If not, hand it over to a developer to make the fixes. Once the fixes are in, always re-run the audit to double-check that everything is in order.
9. Content Management System
I think the best use case for ChatGPT and Claude is actually content management. This has been by far the biggest time-saver for me. Most sites have a CMS—that could be WordPress, Ghost, Contentful, Webflow, Shopify, Squarespace—whatever one you went with. Anyone who’s managed content in a CMS knows how tedious and time-consuming it is. You might spend hours uploading content, adding images, and filling in all the text fields. It’s an extremely boring admin task that used to take hours of my time. Especially when I am working with custom-built systems, they are finicky: they don’t do what they should, they need more features, and they don’t display content the way I want—it’s just a pain in the neck.
Depending on your CMS, you can set up an MCP and offload all of this to Chat or Claude. What I do is basically create a folder on my desktop that matches all the content on the website. Each page is saved as a Markdown file. If I want to add new pages or edit content, all I have to do is open the Markdown file, edit the text, save it, and tell Chat that it’s been updated and to go update the live page. That might not seem like much, but when you are writing a dozen pages, it’s hugely helpful. All you need to do is work on the text and then hand it all off to Chat, who can prefill everything in the CMS ready to publish. You’ll still have to manage images, so that’s not completely taken care of (yet!).
SEO & writing services
Hi 👋 I’m Eli and I help websites grow through SEO & content writing. SEO is one of the best and cheapest ways to grow sustainable online traffic in 2026. If you need help gaining more visibility on Google and building a strong customer acquisition channel, shoot me a message!