Slide 1

Slide 1 text

SEO & AI The Opportunities & The Pitfalls

Slide 2

Slide 2 text

30th November 2022 Chat GPT Launched Backed by Microsoft & Developed by OpenAI 6th February 2023 Google Bard announced With access granted to a small sample of trusted testers February 7th 2023 An AI powered Bing was launched Out of nowhere. Downloading Edge and setting Bing as default are pre-requisites. Bing beat Google to be the first AI assisted search available to the public March 14th 2023 GPT-4 is released Turbo charging ChatGPT etc considerably. More accurate, creative and the ability to serve 25,000 words. Visual inputs also possible. March 16th 2023 Microsoft releases Copilot. Integrating with MS Office and providing Business Chat utlising LLM to enhance 365 apps and personal data such as calendars and emails May 10th 2023 Google souped up Bard and announced new AI powered features. SGE & Perspectives being the most significant, which is still not available in the UK but caused controversy amongst SEOs around privacy organic SERPs May 23rd 2023 Bing announces improvements to Copilot via Windows 11 OS. This was aimed at ensuting Copilot could be a personal assistant, run commands, generate content and retrieve info. Quick recent timeline of AI & search

Slide 3

Slide 3 text

AI & SEO * All images courtesy of artificial intelligence 3

Slide 4

Slide 4 text

What does ChatGPT think… 1. Lack of human touch: While AI can automate various SEO tasks, it may lack the creativity, intuition, and understanding of human users. AI algorithms may struggle to interpret nuanced user intent, leading to less accurate results and potentially missing out on valuable opportunities. 2. Bias and fairness concerns: AI algorithms are only as good as the data they are trained on. If the training data is biased, the AI may perpetuate and amplify those biases, leading to unfair or discriminatory search results. Careful attention must be paid to ensure fairness and address potential biases in AI-driven SEO practices. 3. Dependence on data quality: AI algorithms require large amounts of high-quality data to learn and make accurate predictions. If the input data is incomplete, outdated, or biased, it can negatively impact the performance of AI-driven SEO strategies. Regular data quality checks and continuous improvement efforts are necessary to mitigate this pitfall. Let’s not forget that until March, the data being served via ChatGPT was 2 years old. This has improved, but errors are still very common ChatGPT fail to mention duplication, which is already an issue Source: ChatGPT Prompt: “Describe the pitfalls and opportunities of AI when it comes to SEO”

Slide 5

Slide 5 text

Pessimists would say ChatGPT is a glorified Markov Chain… Chat GPT ChatGPT is a type of language model that is based on the transformer architecture, which is a neural network-based approach to natural language processing (NLP) Markov Chain Markov Chain is a mathematical model that can be used to generate text, it is a relatively simple model compared to neural network- based models like ChatGPT.

Slide 6

Slide 6 text

SEOs are continually look to find ways to spend less time doing grunt work… Automation • Just because a technology or tactic exists and everyone else is doing it, doesn’t mean you need to • But if AI can be used to automate laborious tasks and free up time for more strategic activity then why not SEO Strategy ROI from SEO

Slide 7

Slide 7 text

Resource for implementing technical SEO or copywriting always an issue Python for SEO automation Content Spinning (more on that later) Open AI (long before ChatGPT) • The ability to generate large amounts of content using AI isn’t new and has been possible via OpenAI since 2015. ChatGPT has just made this super accessible, cheap and easy

Slide 8

Slide 8 text

But currently, AI cannot tell you how to…. • Pick the right CMS or platform for your (or your client’s) website • Resonate with your audience and chosen personas • Decide on SEO decisions that require common sense and data that isn’t available to OpenAI • Create content that is guaranteed to boost interaction or engagement metrics, unless heavily QC’ed • Human intuition is still vital

Slide 9

Slide 9 text

ChatGPT and OpenAI in general is only as good as the prompts you give it • We’ve all messed around with ChatGPT for generating content • Finding the best prompts is time consuming and even when you do, this is likely to be very similar to other content on the web • Clients and in house teams have already been falling into the trap of generating weak content and publishing it, unchecked

Slide 10

Slide 10 text

Browser plugins such as AIPRM are game changers

Slide 11

Slide 11 text

Used effectively, they squeeze the very best out of OpenAI

Slide 12

Slide 12 text

SEO crawlers are already being built that harness AI to automate • Automated reporting • Auto archive GSC data beyond 16 months • Never use regex again • Add dynamic tags to keywords for automatic categorization • Sync crawl data with implementation to automatically write dev tickets and send them – and confirm implementation • Real time alerts • Automated internal linking based on bot hits, internal PageRank and relevancy

Slide 13

Slide 13 text

Other tools allow for full onpage optimisation without leaving their platform • Other tools are looking to build the ultimate optimisation suite, allowing for content briefing, keyword clustering, SERP similarity and crawling SERPs in other countries without the need for a VPN

Slide 14

Slide 14 text

This is all possible, due to human intuition • The vast amount of data accessible at ever improving speeds is incredibly valuable but already we are seeing the initiative being taken by humans to get the best out of this data • LLM’s cannot currently do this independently, and when it is that sophisticated then we should all start to worry • Still very much a garbage in, garbage out scenario which puts the onus on us to utilise the technology in the smartest way possible

Slide 15

Slide 15 text

• Using poor prompts to create content that is not engaging and potentially duplicate • Presuming that the output is accurate, or up to date • Encourages lazy SEO, or just work in general • Failing to QC content being generated • Lack of human intuition or common sense • Creating content that does not resonate with your audience and is for search engines (directly against Google’s guidance) Pitfalls Opportunities • Enriching content, scraping data such as PAA and rankings to make recommendations which is then created largely by a human • Automating annoying technical SEO tasks and even dev ticket generation • Content briefing for copywriters based on current rankings, word counts and page structure of most visible websites • Keyword clustering to avoid cannibalisation and duplicate content • Easily establishing entities and topics that search engines are using in their algorithms • Enhanced scraping, and all SEO’s love to scrape!

Slide 16

Slide 16 text

Thanks! 16 Links