Mark Williams-Cook

Me doing an #SEO audit on a site because someone wants “to get an SEO involved before launch“, but you were only contacted a week before the whole thing goes live.

Maybe this is a good way to catch LinkedIn ChatGPT spammers? Prompt injection in my posts?

This is why we halted the AlsoAsked MCP. This is why I have been perpetually warning organisations to only use GenAI where it doesn’t matter if the output is wrong. This is why you can’t automate SEO with AI. The list goes on.

When you see the conference has put you on stage directly after both Aleyda Solís and Lily Ray 😂

That feeling sometimes when you are working with SEO measurement 🤡 iykyk

Unsolicited #SEO tip: Don't bet against Google. We've had a raft of "AI experts" telling us that people are going to stop using Google, and it will all be going to ChatGPT. ⤵️

Apart from the actual reality of traffic, this data from Similarweb shows that OpenAI's GenAI traffic share is quickly being eaten into by Google's Gemini.

While influence may be moving away from company websites and centralised (arguable a bad thing for consumers), I would be surprised to see this change slow down. Google is still highly profitable and OpenAI is reliant on investor funding, which will be a lot less appealing on a downward share trajectory.

⚠️ Free tool: Find intent gaps in your content automatically! I've made an accessible, no API key, no login, no payment required demo of the content improvement workflow I wrote a guide for at IntentGaps[dot]com

What does this tool do?

🖥️ Enter a URL and the tool will scrape that page, read the main content and try and determine the overall page topic by the title/h1. It uses OpenAI to do this, so it isn't perfect, you have the opportunity to edit it if you think it is slightly off.

🔄 Next, the tool will use the AlsoAsked API to fetch the nearest proximity intent questions to your topic. Why? Because the data shows* that sites that answer these questions most thoroughly tend to rank higher.

🤖 Next, we pass your main page content and the AlsoAsked questions to AI to ensure the provided questions are relevant, and then we score the page based on if those questions are fully, partially, or not answered at all.

🚦 You then get a scorecard to give you some easy pointers on where you might be able to improve your content.

I posted a guide** of how to do this at scale on the Search Engine Land in 2024 with Screaming Frog, AlsoAsked and OpenAI, but of course it requires paid software, and paid API keys. This tool is designed as an easy access version to show you what is possible with AlsoAsked data!

In the few cases where there is no PAA tree generated by Google for your topic, I have used OpenAI to generate what it thinks to be some helpful questions to keep things moving.

I built this over the weekend, so I am *sure* something will break or be wonky, and I welcome your feedback in the comments! ⤵️

I critiqued a post/article written by Suganthan Mohanadasan from Keyword Insights / Snippet Digital and his response shows why he is one of the smartest people in the industry ⤵️

He accepts the criticism is of the idea, not the person, he immediately double-checks his thinking, makes updates, and gets a better end result.

I try and exercise this as well (although I am not perfect at it): I want people to criticise my ideas, my thinking, my research, as it makes me better.

There are a lot of reasons I have huge amount of respect for him, but I wanted to highlight this one.

We could all be a bit more like Suganthan Mohanadasan (also follow him).

Jesse Dwyer from Perplexity: "it is currently accurate to say (most) traditional SEO best practices still apply." (In reference to 'GEO'). But maybe he needs a GEO expert to tell him that isn't true? 🙃

Unsolicited #SEO tip: Don't worry about this, nothing has changed. It's likely to do with the previous impressions bug Google had. Carry on with your day 💅

Google has dropped a new guide for "Optimizing your website for generative AI features on Google Search", here is what they said you DON'T NEED:

- llms.txt or AI markdown
- to "chunk" content
- rewrite content for AI systems
- overfocus on structured data

Who would have thought? 🙃

Link in comments to the full guide. Thanks again to Chris Green for the heads up.

I had a wonderful time this week at Athens SEO: such a lovely atmosphere, super organisation (that's the first time I've seen a shuttle bus for speakers), an engaged crowd full of questions and some excellent talks.

Huge thank you firstly to Eva Mermingi, Francesca Kavvadia, and Andreas Georgatsos for your organisational orchestration to pull the whole thing off, and I am humbled that you would invite me to speak.

Secondly, thank you for the kind compliments and thoughtful discussions I had after my talk, it was a pleasure meeting so many new and interesting people. 😎

The first talks I saw were a very flamboyant delivery on Javascript by Martin Splitt, Bengü Sarıca Dinçer following that hugely difficult act with also her own brand of humour and actionable analytics advice, and a deep dive with Simone De Palma on data analysis. In the afternoon I was treated to talks by Jonathan Moore, Judith Lewis, and Yagmur Simsek who handled the late shift perfectly.

Huge fun also catching up with my industry peers and friends Iva Jovanovic, Jamie Indigo, Lily Ray, Aleyda Solís, Dan Taylor, John Campbell, Chris Lever, Frank van Dijk 📈, Dixon Jones and many more.

Hope to see Athens SEO next year!