8. But many of these tips can
unknowingly lead to bad outcomes.
9. Understanding how the tools work
(and don’t work) and applying critical
thinking when using AI is necessary.
10. LLMs are good at:
• Language translation
• Content summarization
• Helping writer's block
• Content generation
• Sentiment analysis
• Question-answering
• Personalization
• Speech recognition
• Stylized writing (applying Poe to x)
• Refurbishing content for social
• Clarifying long or complex content
• Correcting spelling and grammar
LLMs are NOT good at:
• Representing marginalized groups
• Being factual 100% of the time
• Data retrieval (like keyword research)
• Current events
• Predicting the future
• High-level strategy
• Reasoning & Logic
• Understanding or delivering jokes
• Generalization
• Understanding context
• Handling uncommon scenarios
• Emotional intelligence
11. Relying on AI to do the job a human
should (still) be doing (for now)
can cause major problems.
14. These can be subtle inaccuracies that
only a true expert would be able to spot.
16. Seems legit at first glance…
Not confirmed SEO
ranking factors.
17. Of course, SEO isn’t heart surgery…
but what other information are these
tools getting just slightly wrong?
18. “Large language models will tell you they
can do anything.
But for many things there's much better
and more accurate tools (like python, R or
even Excel.)
Using an LLM for these tasks is like
hammering nails with a screwdriver.”
– Ryan Jones @ryanjones
25. Training cut off September 2021
Browse with Bing disabled
Some plugins can access internet (Webpilot)
Can access real-time information
via Bing’s index
Can access real-time information
Training cut off December 2022
But knows some 2023 information
37. According to Google, AI content can be ok if:
01
IT IS original
& high-quality
IT DEMONSTRATES
E-E-A-T
02
IT SATISFIES
the Helpful Content
System
03
IT DOESN’T
propagate
misinformation
04
IT ISN’T created with
the primary purpose of
manipulating search results
05
38. However, there are 3 important caveats:
Google recommends
including author
names (for some
content) Google
recommends
disclosing use of AI
Google
recommends
against listing AI
as the “author” of
the content
43. Let’s review some fundamentals of E-E-A-T:
• Experience, Expertise, Authoritativeness and Trustworthiness
• Not a ranking factor, comes from Search Quality Guidelines
• Google doesn’t tell us what components of E-E-A-T play a role in
ranking, except links and PageRank
•
Matters more for YMYL (your money, your life) sites
56. Using AI tools to compare pages and
improve their information gain scores
57. An information gain score… indicates
“additional information included by a
page beyond the information
contained in other pages already
presented to the user.”
Bill Slawski
2020 Google Patent:
Contextual Estimation of
Link Information Gain
61. Once URLs have been analyzed, you can ask:
• What content is included on page A that is not
on page B? (information gain)
• Do the pages reflect different levels of
experience, expertise, authoritativeness and
trustworthiness? If so, why?
• How could page A be improved to have more
comprehensive information than page B?
80. Ask ChatGPT to
create new
headlines based
on your top-
performing
headlines
Blurred for client confidentiality
81. Ask ChatGPT to
create a framework
for you to follow in a
table format, based
on the top
performing
headlines
(Generic prompt shown – ask it to base
the headlines on your top performers)
84. Used Google patent
about classifying
websites (website
representation vectors)
Analyze top-performing
medical content against
the patent
Ask how an article
compares and can be
improved
97. Process:
• Collect frequently asked questions about your brand
• Answer questions clearly and concisely on your site(s)
in FAQs, About page, homepage, etc.
• Support facts with structured data (you can use AI
tools to generate)
• Encourage search engine indexing (Search Console,
IndexNow)
• Monitor SGE & Bing Chat results