- Oct 9, 2013
- 3,471
- 14,481
I've got an AIO/GEO one in the pipeline.
I took the stance of that being a scam a year ago, and it was back then, BUT, things have really changed.
It's becoming a significant source of traffic now, and with the release of Atlas, things are going to get very interesting.
But other than that.. What would people like to see? I'll do them ALL when I have time, but feedback is helpful.
Choose one in the poll, and reply in the thread too. Tell me if you'd like to see multiple, or if there's any I've missed.
Tell me about some of the challenges you're facing in SEO today.
What kind of projects are people doing?
Blogging/info is very dead these days. At least PURE info sites unless you have deep pockets and can build up authority sites. Legit DR 60+'s
AIO/GEO - How to get more clicks from chatgpt, perplexity, gemini, google AI overviews
There are some distinctive ways to get more traffic from AIs. The type of content you need to make now is actually very different to what you make for Google.
You have 2 types. Content you make to rank on Google, and content you make for AIO.
Google's Legitimacy Filter - Why some sites don't rank and how to fix it
This is a new term I've coined called a "legitimacy filter". It is absolutely something that exists within the Google algorithm.
In fact, THIS is what the helpful content updates were.
It was the perfecting of the legitimacy filter.
Google can't tell good from bad content. They never could. What they CAN do is determine if you are legitimate or not.
All your sites MUST pass this otherwise you won't rank.
Something I've also observed is that there seems to be a permanent RED switch with regards to this.
Here's a diagram
A brand new site lands in yellow. The more signals of legitimacy you have, the more you move up. Hit green, and your site is unlocked.
BUT, if you do the opposite, and have NEGATIVE signals, without positive, you can pull yourself down into a red flag, permanent disable.
This is why you get sites that never recovered from the helpful content update no matter what they do.
They likely did this because.. Think about it. Statistically, if one day, a site has a TON of spam signals, but after an update, it then "improves". What are the chances it's really legit vs the chances it's faking it?
This is how Google engineers likely think. If I think this(I have an engineering mind. I think in systems/probabilities/connections). Chances are they probably did too.
It's also possible they use this filter when it comes to backlinks counting too.
If you are high green, more backlinks will count.
If you're low green, only more legit looking backlinks count. <-- This is pure speculation, but it makes sense.
They would naturally want to favor sites that have more legitimacy signals.
You can see this in the serps today. Big brands dominate as through many points in the SEO timeline, but you also have many small, topically relevant sites that do EXTREMELY well, and if you look closely at them, you can clearly see they are legit businesses. They have real people behind them.
Sites that have that same topical relevance but have nothing connecting that site to real entities in the real world don't do as well. You rarely see sites today ranking that don't have a visible team/company behind them.
In the past, you'd have sites with dense topical relevance, that would rank with just a thin about page, privacy policy and terms of service. Today. I don't see this.
It's also worth noting that, while you're in yellow, you will still rank, but you are in a sandbox of sorts. You are throttled. There's some sort of limit. They may also be testing the pages on your site for user feedback too at this stage, and bad user feedback could put you in red, while good feedback without other signals will just keep you in yellow.
This guide will be pretty hot. Pretty much novel information in the SEO industry and WITH techniques you can use to grade your own site. Not just theory!
Composition Theory of Ranking - Why all on-page tools in 2025 like surfer and POP dont work in 2025
Surfer, POP and all the other on-page tools, even entity based tools like inlinks work sometimes, and not others.
The problem is they're missing something and making assumptions that for you is like rolling a dice.
You might get lucky and hit a 6 and the advice will work, or you might not and never know why it works sometimes and not others.
My composition theory of ranking will explain WHY they don't always work, and WHAT you can do instead.
It also lines up with my reiterated statements "You Don't Rank Pages with Links" that I love to keep reminding everyone of.
You'll understand why some pages rank when you link to them, and others don't, and why sometimes when you link to page A, page B goes up but page A doesn't.
Google HAVE ALWAYS used PageRank. It exists in 2025. It's at the core of everything.
The only difference is there's endless filters. You have to get through them all in order to get your juicy PR reward.
And how does PR work? It is called Page Rank. It passes through pages and increases the score of each page.
If you link to page A, then page A gets a PR boost. Page A then links to B, C, D and E, and they get a % PR boost(less, but still substantial). Then they link on, and it goes round the whole site.
Another thing to mention is, PR passes based on contextual relevance. It doesn't pass based on SITE RELEVANCE, which is why you don't need links from sites in your niche. It passes based on the page relevance, or rather, the contextual relevance surrounding the link. Yet another filter!!
How to find 100's of low comp keywords for e-commerce sites
This one is related to the composition theory of ranking.
Technically I'll have to write about it before this one.
This shows you how to use the composition theory of ranking to identify keywords you CAN rank for.
Little hint there in the wording
Forget about 'low comp keywords'. THEY DONT EXIST IN 2025!!
There's no such thing as high or low comp!
There are keywords you CAN rank for and keywords you CANT.
Within that, there are difficulty grades.
Old style: Low comp, medium comp, high comp.
New style: Pre-filter: Can I rank? How many spaces is available? What's the difficulty level on page 1.
So 3 things.
First, a pre-filter
Second, there might only be 3 positions you CAN rank for.
Or there might be 6.
Then how strong are those 3 or 6?
3 and low comp?
6 and low comp? -- Even better
6 and medium comp - Decent, and do-able.
3 and high comp? You could, but you'll need deep pockets.
Google Webmaster Tools(Search Console) Masterclass
Awesome things you can do with webmaster tools to improve your site.
How to make Claude Code your Director of SEO
In this guide(I'll maybe make a video to accompany it since it'll be more hands on) I'll show you how to turn Claude Code into your very own SEO expert to help you diagnose problems with your site, analyse competitors and help you make better decisions.
Cannibalisation - The silent SEO killer
A guide to finding and fixing cannibalisation on your site.
Do your own Technical SEO Audit with AI
A super practical, no BS, no theory-heavy, ultra-comprehensive guide to doing your own technical SEO.
(so many -'s. People are going to think I'm chatgpt)
Why even bother doing this you might ask? Agencies will do this for you..
The advantage to doing this yourself is you can use it as a tool to pre-screen agencies.
Create your own technical SEO report, then get 5 potential agencies to do one for you.
Pop them all into your favorite LLM and do this:
NOTE: You cannot ask the AI to compare your audit report with 5 others. IT WILL NOT DO A GOOD JOB!
You must use an orchestrator model framework for this.
This is how:
Open up 5 chat windows with an LLM.
In each say:
"Here is a master technical SEO audit I did myself using AI after following instructions from an industry SEO expert. Consider this the source of truth for my site xxx.com. I found an SEO agency that did their own audit on my site. I have attached both their audit and the master SEO audit. Please compare and contrast and give me your opinion on how close they are to the master audit. Did they miss anything? If so, what. Consider this a yellow flag. Bad, but not a game-ender, as they may just be trying to save time doing a quick audit. Do they have CONTRADICTORY information in their report? Consider this a RED flag. Give me a detailed comparison report, grading each part as green(caught the issue from the master) or red(information contradicts master). You can also have a special 'blue' grade for points that the agency's audit covered, but the master missed. (However, note that this would be harder to verify as it may be something either superfluous or they are actually highlighting an issue that is not a problem. Use your own intelligence and knowledge of SEO's to make an assessment about the blues) Additionally have a report for yellows(missed information that was in the master) as well as a total grading, number of greens, yellows and reds, as well as total practical action points in the audit, and give them a score out of 10. Consider the master to be a 8/10 audit. 9 or 10 would surpass it. For you to grade it 9 or 10 you would need to be exceptionally confident that the blues you have found are indeed actual valid issues."
THEN.
Take the output from all 5. Put them in text files.
Go to the model again, this time it's your orchestrator model.
And say
"I have a master technical SEO audit created for my site xxx.com which I created myself, based on a guide from an industry SEO expert. Consider this our source of truth. I then got 5 SEO agencies to do technical SEO audit. You are an orchestrator model and your job is to provide the final assessment of the quality of each agency's audit.
I have used 5 separate LLMs to each individually assess 1 agency report.
This is the prompt I used
"Here is a master technical SEO audit I did myself using AI after following instructions from an industry SEO expert. Consider this the source of truth for my site xxx.com. I found an SEO agency that did their own audit on my site. I have attached both their audit and the master SEO audit. Please compare and contrast and give me your opinion on how close they are to the master audit. Did they miss anything? If so, what. Consider this a yellow flag. Bad, but not a game-ender, as they may just be trying to save time doing a quick audit. Do they have CONTRADICTORY information in their report? Consider this a RED flag. Give me a detailed comparison report, grading each part as green(caught the issue from the master) or red(information contradicts master). You can also have a special 'blue' grade for points that the agency's audit covered, but the master missed. (However, note that this would be harder to verify as it may be something either superfluous or they are actually highlighting an issue that is not a problem. Use your own intelligence and knowledge of SEO's to make an assessment about the blues) Additionally have a report for yellows(missed information that was in the master) as well as a total grading, number of greens, yellows and reds, as well as total practical action points in the audit, and give them a score out of 10. Consider the master to be a 8/10 audit. 9 or 10 would surpass it. For you to grade it 9 or 10 you would need to be exceptionally confident that the blues you have found are indeed actual valid issues."
I have attached the master report, named master.txt, the 5 agency reports, named agency_report_1.txt to agency_report_5.txt, as well as the 5 sub-agent reports, named sub_agent_report_1.txt to sub_agent_report_5.txt
Your job is to assess each of the agency audits and read the sub-agent reports for that agency. You will:
1) Confirm the sub-agent's work and identify any errors they have made or items they have missed.
2) Grade each agency out of 10, and also note the sub-agent's grade beside. If yours differs, explain why.
3) Make your recommendation for which agency you would recommend based on these audits.
"
For models(if you have claude pro and chatgpt pro) use 5 x deep thinking gpt 5's for sub-agents and opus 4.1 for the orchestrator.
If you only have plus, then I the highest think you can in chatgpt. If you lack claude at all, then use the deepest thinking gpt5 you have.
( gpt5 deep think tends to do better for most tasks than gpt 5 pro. Pro is better for highly technical and scientific tasks. For this deep will be better, but feel free to test both or other variations!!)
I took the stance of that being a scam a year ago, and it was back then, BUT, things have really changed.
It's becoming a significant source of traffic now, and with the release of Atlas, things are going to get very interesting.
But other than that.. What would people like to see? I'll do them ALL when I have time, but feedback is helpful.
Choose one in the poll, and reply in the thread too. Tell me if you'd like to see multiple, or if there's any I've missed.
Tell me about some of the challenges you're facing in SEO today.
What kind of projects are people doing?
Blogging/info is very dead these days. At least PURE info sites unless you have deep pockets and can build up authority sites. Legit DR 60+'s
AIO/GEO - How to get more clicks from chatgpt, perplexity, gemini, google AI overviews
There are some distinctive ways to get more traffic from AIs. The type of content you need to make now is actually very different to what you make for Google.
You have 2 types. Content you make to rank on Google, and content you make for AIO.
Google's Legitimacy Filter - Why some sites don't rank and how to fix it
This is a new term I've coined called a "legitimacy filter". It is absolutely something that exists within the Google algorithm.
In fact, THIS is what the helpful content updates were.
It was the perfecting of the legitimacy filter.
Google can't tell good from bad content. They never could. What they CAN do is determine if you are legitimate or not.
All your sites MUST pass this otherwise you won't rank.
Something I've also observed is that there seems to be a permanent RED switch with regards to this.
Here's a diagram
A brand new site lands in yellow. The more signals of legitimacy you have, the more you move up. Hit green, and your site is unlocked.
BUT, if you do the opposite, and have NEGATIVE signals, without positive, you can pull yourself down into a red flag, permanent disable.
This is why you get sites that never recovered from the helpful content update no matter what they do.
They likely did this because.. Think about it. Statistically, if one day, a site has a TON of spam signals, but after an update, it then "improves". What are the chances it's really legit vs the chances it's faking it?
This is how Google engineers likely think. If I think this(I have an engineering mind. I think in systems/probabilities/connections). Chances are they probably did too.
It's also possible they use this filter when it comes to backlinks counting too.
If you are high green, more backlinks will count.
If you're low green, only more legit looking backlinks count. <-- This is pure speculation, but it makes sense.
They would naturally want to favor sites that have more legitimacy signals.
You can see this in the serps today. Big brands dominate as through many points in the SEO timeline, but you also have many small, topically relevant sites that do EXTREMELY well, and if you look closely at them, you can clearly see they are legit businesses. They have real people behind them.
Sites that have that same topical relevance but have nothing connecting that site to real entities in the real world don't do as well. You rarely see sites today ranking that don't have a visible team/company behind them.
In the past, you'd have sites with dense topical relevance, that would rank with just a thin about page, privacy policy and terms of service. Today. I don't see this.
It's also worth noting that, while you're in yellow, you will still rank, but you are in a sandbox of sorts. You are throttled. There's some sort of limit. They may also be testing the pages on your site for user feedback too at this stage, and bad user feedback could put you in red, while good feedback without other signals will just keep you in yellow.
This guide will be pretty hot. Pretty much novel information in the SEO industry and WITH techniques you can use to grade your own site. Not just theory!
Composition Theory of Ranking - Why all on-page tools in 2025 like surfer and POP dont work in 2025
Surfer, POP and all the other on-page tools, even entity based tools like inlinks work sometimes, and not others.
The problem is they're missing something and making assumptions that for you is like rolling a dice.
You might get lucky and hit a 6 and the advice will work, or you might not and never know why it works sometimes and not others.
My composition theory of ranking will explain WHY they don't always work, and WHAT you can do instead.
It also lines up with my reiterated statements "You Don't Rank Pages with Links" that I love to keep reminding everyone of.
You'll understand why some pages rank when you link to them, and others don't, and why sometimes when you link to page A, page B goes up but page A doesn't.
Google HAVE ALWAYS used PageRank. It exists in 2025. It's at the core of everything.
The only difference is there's endless filters. You have to get through them all in order to get your juicy PR reward.
And how does PR work? It is called Page Rank. It passes through pages and increases the score of each page.
If you link to page A, then page A gets a PR boost. Page A then links to B, C, D and E, and they get a % PR boost(less, but still substantial). Then they link on, and it goes round the whole site.
Another thing to mention is, PR passes based on contextual relevance. It doesn't pass based on SITE RELEVANCE, which is why you don't need links from sites in your niche. It passes based on the page relevance, or rather, the contextual relevance surrounding the link. Yet another filter!!
How to find 100's of low comp keywords for e-commerce sites
This one is related to the composition theory of ranking.
Technically I'll have to write about it before this one.
This shows you how to use the composition theory of ranking to identify keywords you CAN rank for.
Little hint there in the wording
Forget about 'low comp keywords'. THEY DONT EXIST IN 2025!!
There's no such thing as high or low comp!
There are keywords you CAN rank for and keywords you CANT.
Within that, there are difficulty grades.
Old style: Low comp, medium comp, high comp.
New style: Pre-filter: Can I rank? How many spaces is available? What's the difficulty level on page 1.
So 3 things.
First, a pre-filter
Second, there might only be 3 positions you CAN rank for.
Or there might be 6.
Then how strong are those 3 or 6?
3 and low comp?
6 and low comp? -- Even better
6 and medium comp - Decent, and do-able.
3 and high comp? You could, but you'll need deep pockets.
Google Webmaster Tools(Search Console) Masterclass
Awesome things you can do with webmaster tools to improve your site.
How to make Claude Code your Director of SEO
In this guide(I'll maybe make a video to accompany it since it'll be more hands on) I'll show you how to turn Claude Code into your very own SEO expert to help you diagnose problems with your site, analyse competitors and help you make better decisions.
Cannibalisation - The silent SEO killer
A guide to finding and fixing cannibalisation on your site.
Do your own Technical SEO Audit with AI
A super practical, no BS, no theory-heavy, ultra-comprehensive guide to doing your own technical SEO.
(so many -'s. People are going to think I'm chatgpt)
Why even bother doing this you might ask? Agencies will do this for you..
The advantage to doing this yourself is you can use it as a tool to pre-screen agencies.
Create your own technical SEO report, then get 5 potential agencies to do one for you.
Pop them all into your favorite LLM and do this:
NOTE: You cannot ask the AI to compare your audit report with 5 others. IT WILL NOT DO A GOOD JOB!
You must use an orchestrator model framework for this.
This is how:
Open up 5 chat windows with an LLM.
In each say:
"Here is a master technical SEO audit I did myself using AI after following instructions from an industry SEO expert. Consider this the source of truth for my site xxx.com. I found an SEO agency that did their own audit on my site. I have attached both their audit and the master SEO audit. Please compare and contrast and give me your opinion on how close they are to the master audit. Did they miss anything? If so, what. Consider this a yellow flag. Bad, but not a game-ender, as they may just be trying to save time doing a quick audit. Do they have CONTRADICTORY information in their report? Consider this a RED flag. Give me a detailed comparison report, grading each part as green(caught the issue from the master) or red(information contradicts master). You can also have a special 'blue' grade for points that the agency's audit covered, but the master missed. (However, note that this would be harder to verify as it may be something either superfluous or they are actually highlighting an issue that is not a problem. Use your own intelligence and knowledge of SEO's to make an assessment about the blues) Additionally have a report for yellows(missed information that was in the master) as well as a total grading, number of greens, yellows and reds, as well as total practical action points in the audit, and give them a score out of 10. Consider the master to be a 8/10 audit. 9 or 10 would surpass it. For you to grade it 9 or 10 you would need to be exceptionally confident that the blues you have found are indeed actual valid issues."
THEN.
Take the output from all 5. Put them in text files.
Go to the model again, this time it's your orchestrator model.
And say
"I have a master technical SEO audit created for my site xxx.com which I created myself, based on a guide from an industry SEO expert. Consider this our source of truth. I then got 5 SEO agencies to do technical SEO audit. You are an orchestrator model and your job is to provide the final assessment of the quality of each agency's audit.
I have used 5 separate LLMs to each individually assess 1 agency report.
This is the prompt I used
"Here is a master technical SEO audit I did myself using AI after following instructions from an industry SEO expert. Consider this the source of truth for my site xxx.com. I found an SEO agency that did their own audit on my site. I have attached both their audit and the master SEO audit. Please compare and contrast and give me your opinion on how close they are to the master audit. Did they miss anything? If so, what. Consider this a yellow flag. Bad, but not a game-ender, as they may just be trying to save time doing a quick audit. Do they have CONTRADICTORY information in their report? Consider this a RED flag. Give me a detailed comparison report, grading each part as green(caught the issue from the master) or red(information contradicts master). You can also have a special 'blue' grade for points that the agency's audit covered, but the master missed. (However, note that this would be harder to verify as it may be something either superfluous or they are actually highlighting an issue that is not a problem. Use your own intelligence and knowledge of SEO's to make an assessment about the blues) Additionally have a report for yellows(missed information that was in the master) as well as a total grading, number of greens, yellows and reds, as well as total practical action points in the audit, and give them a score out of 10. Consider the master to be a 8/10 audit. 9 or 10 would surpass it. For you to grade it 9 or 10 you would need to be exceptionally confident that the blues you have found are indeed actual valid issues."
I have attached the master report, named master.txt, the 5 agency reports, named agency_report_1.txt to agency_report_5.txt, as well as the 5 sub-agent reports, named sub_agent_report_1.txt to sub_agent_report_5.txt
Your job is to assess each of the agency audits and read the sub-agent reports for that agency. You will:
1) Confirm the sub-agent's work and identify any errors they have made or items they have missed.
2) Grade each agency out of 10, and also note the sub-agent's grade beside. If yours differs, explain why.
3) Make your recommendation for which agency you would recommend based on these audits.
"
For models(if you have claude pro and chatgpt pro) use 5 x deep thinking gpt 5's for sub-agents and opus 4.1 for the orchestrator.
If you only have plus, then I the highest think you can in chatgpt. If you lack claude at all, then use the deepest thinking gpt5 you have.
( gpt5 deep think tends to do better for most tasks than gpt 5 pro. Pro is better for highly technical and scientific tasks. For this deep will be better, but feel free to test both or other variations!!)
