From Search Rankings to Answer Synthesis - How to Weather The AI Storm

Slide 1

Slide 1 text

From Search Ranking to Answer Synthesis: How to Weather the AI Storm Dawn Anderson - Bertey

Slide 2

Slide 2 text

Who's Dawn Anderson?

Slide 3

Slide 3 text

2023 - The Dawn of Generative Information Retrieval

Slide 4

Slide 4 text

In Plain Language - From 10 Blue Links... To AI Overviews...

Slide 5

Slide 5 text

The Dawn of Generative Information Retrieval • Represents the biggest change in information retrieval in 70 years • Yes, since before the web and as far back as post Second World War days

Slide 6

Slide 6 text

SIGIR Generative IR Workshop 2023 v Google Search Central NYC 2025 – Both Show Generative IR Becomes The Core of Search

Slide 7

Slide 7 text

In the world of Information Retrieval There's Been Trouble At t'mill

Slide 8

Slide 8 text

Generative Information Retrieval is Google's 'Rush Job' • In response to Open AI / ChatGPT • To align with Broder's 'Delphic Costs' 'effortless search' concept • 'Minimum Viable Product' with Reinforcement Learning From Human Feedback (RLHF) • Hallucinations galore

Slide 9

Slide 9 text

Generative IR is NOT Robust. Not Source Faithful. Not Factual

Slide 10

Slide 10 text

"Involution Instead of Evolution" (Ricardo Baeza- Yates)

Slide 11

Slide 11 text

Bender & Shah (2024) refer to the web as a "potential endangered information ecosystem" • Source: Envisioning Information Access Systems: What Makes for Good Tools and a Healthy Web? Bender & Shah, 2024

Slide 12

Slide 12 text

Generative IR Does Not Meet The Various Types of Belkin Information Seeking Strategies (ISS) (Bender & Shah, 2024)

Slide 13

Slide 13 text

Nevertheless... Generative IR Rumbles Onwards

Slide 14

Slide 14 text

So...What Should SEOs Be Doing Right Now About All of This? Be realistic 01 Be forward- thinking 02 Be more consistent than ever before 03 Be technically sound 04 Know thy audience & personas 05 Begin measurement, monitoring and tracking plans 06

Slide 15

Slide 15 text

Be Realistic

Slide 16

Slide 16 text

No content

Slide 17

Slide 17 text

AI Search Overall is Largely Inaccurate https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines- theyre-all-bad-at-citing-news.php

Slide 18

Slide 18 text

You Don't Want to Be Like 'Chicken Licken' AKA 'Henny Penny' here in the U.S.

Slide 19

Slide 19 text

Be Forward Thinking

Slide 20

Slide 20 text

AI Overviews / Generative IR is Rising... Fast

Slide 21

Slide 21 text

This is a Take Over

Slide 22

Slide 22 text

All Google Generative IR / Gen AI / AI Overviews Are Rising Massively

Slide 23

Slide 23 text

Studies by Salt Agency finds LLM traffic converts at a similar rate to organic traffic Source: https://salt.agency/blog/do-users-really- show-higher-intent-when-they-click-through-from- an-llm-to-a-website/

Slide 24

Slide 24 text

Be More Consistent Than EVER Before

Slide 25

Slide 25 text

Two Types of Generative IR Closed book Information comes only from within the model Open book Information comes from within the model and also can come from external sources to provide grouding

Slide 26

Slide 26 text

We are dealing with 'Open Book' Generative IR The gaps get filled with grounding via Retrieval Augmented Generation (RAG)

Slide 27

Slide 27 text

Open Book = Retrieval Augmented Generation (RAG) Helps to Bridge Gaps

Slide 28

Slide 28 text

Data Imputation – Bridging Gaps in Data

Slide 29

Slide 29 text

Disambiguating Equiprobability

Slide 30

Slide 30 text

Disambiguating Equiprobability... But this time it's for LLMs / Generative IR / AI Overviews MAXIMUM LIKELIHOOD ESTIMATION

Slide 31

Slide 31 text

(Probably akin to) Estimation- Maximisation Algorithm Image Source: https://www.geeksforgeeks.org/ml-expectation-maximization-algorithm/

Slide 32

Slide 32 text

Website Homonymy – hat tip -> Dixon Jones

Slide 33

Slide 33 text

Context Works Wonders

Slide 34

Slide 34 text

There's Hope For Schema Yet

Slide 35

Slide 35 text

Add That Context SEOs Relatedness Schema Semantic headings Remove ambiguity Consistency in clustering Clear topic chunking in long form content

Slide 36

Slide 36 text

Be Technically Solid

Slide 37

Slide 37 text

If You Block Google-Extended You Are Not a Candidate For Grounded Search Result Inclusion

Slide 38

Slide 38 text

Avoid (even more so) client side rendered Javascript

Slide 39

Slide 39 text

Although... • Both Googlebot and Google- Extended (Gemini) use WRS (Web Rendering Service) • Google AI bot renders javascript Source: https://www.seroundtable.com/googles-ai-crawler-renders-javascript-39353.html

Slide 40

Slide 40 text

LLMS.TXT - Open for Debate Currently As to its Necessity

Slide 41

Slide 41 text

Are LLMS.txt Akin to the Meta Keywords Tag for Google? https://www.reddit.com/r/TechSEO/comments/1k0kcx9/llmtxt_where_are_we_at/

Slide 42

Slide 42 text

Uptake is Very Low Currently Source: https://www.chris-green.net/post/million-websites-in-search-of-llms-txt

Slide 43

Slide 43 text

Know Thy Audience and Personas

Slide 44

Slide 44 text

Memories & Personalisation

Slide 45

Slide 45 text

Hobbies & Interests

Slide 46

Slide 46 text

Begin Measurement, Monitoring and Tracking Plans

Slide 47

Slide 47 text

Claim that AI Search referral traffic as SEO There will 100% need to be bridges crossed by SEO professionals to maximise LLM and AI Search traffic

Slide 48

Slide 48 text

GA4 Filters to Create

Slide 49

Slide 49 text

No content

Slide 50

Slide 50 text

No content

Slide 51

Slide 51 text

It's An Exciting Time for Information Retrieval & SEO • "Generative question answering is not the end of classical information retrieval, but rather opens up a new frontier of exciting and important research problems." (Najork - SIGIR 2023)

Slide 52

Slide 52 text

Takeaways Be • Be realistic - Don't jump the gun Be • Be forward thinking. Prepare the ground. Increase • Increase consistency assurance across content, entities, schema, data imputation Know • Know thy audience and personas Begin • Begin measurement in traffic (GA) and visibility in AI search (including Google AI Overviews) Be • Be technically sound

Slide 53

Slide 53 text

Thank you! X - @dawnieando Bluesky - @dawnieando Linkedin – Ms Dawn Anderson Bertey.com