Slide 1

Slide 1 text

CONFIDENTIAL Context is the Analytics King? Tech SEO Connect 1

Slide 2

Slide 2 text

JR Oakes 2 JR Oakes is the VP of Strategy for LOCOMOTIVE Agency. He has been an SEO since 2011 and was formerly an architectural glass artist. His focus areas are in SEO, machine learning, language, and user experience. jroakes locomotive.agency

Slide 3

Slide 3 text

CONFIDENTIAL My Interests Have Broadened a Bit 3 Causal ML Generative AI Personalization Engines Technical SEO

Slide 4

Slide 4 text

CONFIDENTIAL CONFIDENTIAL Interestingness 4 Section 2

Slide 5

Slide 5 text

CONFIDENTIAL Efficient Infinite Context Transformers with Infini-attention 5 Source

Slide 6

Slide 6 text

CONFIDENTIAL Mixed Bag 6

Slide 7

Slide 7 text

CONFIDENTIAL Better Model 7 Brand Expertise Trust Transparency Engagement Real Experience

Slide 8

Slide 8 text

CONFIDENTIAL LLMs Love the Word Delve 8 Source

Slide 9

Slide 9 text

CONFIDENTIAL 9 Source

Slide 10

Slide 10 text

CONFIDENTIAL Top-10 Best SEO Agencies* * According to ChatGPT models 10

Slide 11

Slide 11 text

CONFIDENTIAL Top-10 Best AI Writing Tools* * According to ChatGPT models 11 Shocking how consistent this is

Slide 12

Slide 12 text

CONFIDENTIAL Reasons LLM Brand Tracking is Messy 12 Temperature Memory/Projects Random Seeds RAG Search Slight differences in wording Different Models

Slide 13

Slide 13 text

CONFIDENTIAL LibreChat 13

Slide 14

Slide 14 text

CONFIDENTIAL CONFIDENTIAL Trends 14 Section 1

Slide 15

Slide 15 text

CONFIDENTIAL Watching Emerging Trends 15

Slide 16

Slide 16 text

CONFIDENTIAL 16 People are looking at the problems with AI, but not the possibilities.

Slide 17

Slide 17 text

CONFIDENTIAL 17 OR

Slide 18

Slide 18 text

CONFIDENTIAL 18

Slide 19

Slide 19 text

CONFIDENTIAL 19 AI Everyone

Slide 20

Slide 20 text

CONFIDENTIAL Causal Reasoning and Large Language Models “…we find that LLMs can generate text corresponding to correct causal arguments with high probability, surpassing the best-performing existing methods” 20 Source

Slide 21

Slide 21 text

CONFIDENTIAL ■ LLMs can generate causal graphs from metadata alone ■ Capable of counterfactual reasoning in natural language ■ Determine necessary/sufficient causes from text descriptions ■ Capture relevant context and common sense for causal judgments ■ Complement existing statistical causal methods ■ Automate parts of causal analysis previously requiring human experts ■ Enable flexible, natural language interaction for causal tasks 21 Source Causal Reasoning and Large Language Models

Slide 22

Slide 22 text

CONFIDENTIAL CONFIDENTIAL We Code 22 Section 3

Slide 23

Slide 23 text

CONFIDENTIAL 23 This Past Year We Have Been Building.. ■ Built a tool called Npath, that turns GA4 sequence data into user cohorts, ideals paths, and insights. ■ Built a tool that takes a company’s ICP information and turns it into tens of thousands of categorized keywords with competitive metrics. ■ Formalized all competitive gaps to aggregate data based on subjects rather than keywords. ■ Finalizing a tool to detect anomalies buried in thousands of URLs and segments of URLs.

Slide 24

Slide 24 text

CONFIDENTIAL 24 I had a thought

Slide 25

Slide 25 text

CONFIDENTIAL Can we just throw URL page data into a big bucket each month and have LLMs compare and share insights on it? 25

Slide 26

Slide 26 text

CONFIDENTIAL 26 A Cloud Robot Looking for Errors in My URLs

Slide 27

Slide 27 text

CONFIDENTIAL 27 We Need to Put Some Pieces Together

Slide 28

Slide 28 text

CONFIDENTIAL Can we fit all the data and have a dependable output? 28

Slide 29

Slide 29 text

Gemini 29 - 2M Tokens - Schema specification - MIME type specification (e.g. application/json)

Slide 30

Slide 30 text

Tech Debt 30 You kinda get into dependency hell with LLMs. Will it work if they upgrade the model?

Slide 31

Slide 31 text

CONFIDENTIAL Can we afford it? 31

Slide 32

Slide 32 text

CONFIDENTIAL 32 Gemini API Costs Input Pricing $1.25 / 1 million tokens Output Pricing $5.00 / 1 million tokens 1M tokens = ~130 pages 1M tokens = ~700 pages

Slide 33

Slide 33 text

CONFIDENTIAL What if it is not perfect? 33

Slide 34

Slide 34 text

CONFIDENTIAL 34 Models Will Get Better The pace of improvement compared with human- level performance continues to expand across a wide-range of tasks. Source

Slide 35

Slide 35 text

CONFIDENTIAL 35 And More Accurate Models continue to get better at controlling hallucinations. Source

Slide 36

Slide 36 text

CONFIDENTIAL 36 It’s Yours, Change it This is open source, so tweak the prompts your your needs.

Slide 37

Slide 37 text

CONFIDENTIAL Can LLMs do math? 37

Slide 38

Slide 38 text

Ok. There Are Issues Here…. 38 …position improve from 7.85 to 9, an increase of 92.31%...

Slide 39

Slide 39 text

We Have to Be a Bit Creative 39 Just provide current and prior values. We calculate the rest.

Slide 40

Slide 40 text

CONFIDENTIAL What context can we provide? 40

Slide 41

Slide 41 text

CONFIDENTIAL 41 Metrics Content Analysis: ● Clean Content ● Metadata ● Word Count ● Heading Structure ● Image Count ● Schema Markup Page Speed Insights: ● Largest Contentful Paint ● Cumulative Layout Shift ● Interaction to Next Paint ● First Contentful Paint ● Time to Interactive ● Speed Index ● Total Blocking Time ● Performance Score ● First Meaningful Paint Search Engine Performance: ● Clicks (from Google Search Console) ● Impressions (from Google Search Console) ● Click-Through Rate (CTR) ● Average Position (in search results) ● Ranking Keywords ● Top No-Click Queries

Slide 42

Slide 42 text

CONFIDENTIAL 42 Metrics Traffic and User Behavior: ● Organic Sessions ● Organic Users ● Organic New Users ● Bounce Rate ● Average Time on Page ● Engagement Rate ● Revenue ● User Demographics ● Device Categories ● Pages Visited Prior ● Pages Visited Next ● Referring Sites

Slide 43

Slide 43 text

CONFIDENTIAL CONFIDENTIAL SEODP 43 Section 4

Slide 44

Slide 44 text

CONFIDENTIAL An eye in the cloud watching my content 44

Slide 45

Slide 45 text

CONFIDENTIAL 45 SEO Data Platform on Github

Slide 46

Slide 46 text

Requirements 46 1. A developer :-) 2. A few API keys (most are free or very cheap)

Slide 47

Slide 47 text

Modular 47 1. Very clean and extensible extractor workflow. 2. Intuitive to find and update code.

Slide 48

Slide 48 text

Customizable 48 You tell it what to report on.

Slide 49

Slide 49 text

Output 49 Insights in your inbox. Find it here: SEODP

Slide 50

Slide 50 text

Examples 50

Slide 51

Slide 51 text

Examples 51

Slide 52

Slide 52 text

Bonus 52 pip install repocoder

Slide 53

Slide 53 text

CONFIDENTIAL CONFIDENTIAL The End 53 Section null