SEOloger č. 71: Michal Suski: From AI to Authority: Why Al-Generated Content Isn’t Created Equal

Slide 1

Slide 1 text

From AI to Authority: Why AI-Generated Content Isn’t Created Equal Michał Suski Head of Innovation @Surfer

Slide 2

Slide 2 text

Bio ● Master of Science at Forestry ● In IT since 2014 ● SEO since 2016 ● Skiing addict Head of Innovation @Surfer Michał Suski 󰐤 aka dr Surfer

Slide 3

Slide 3 text

17 000+ CUSTOMERS 180+ COUNTRIES 130 000 ACTIVE USERS TRUSTED BY

Slide 4

Slide 4 text

Agenda 1. Generative AI main struggle 2. Ingredients for growth 3. Case studies

Slide 5

Slide 5 text

Stoic graﬃti Looking ok, isn’t it?

Slide 6

Slide 6 text

Miedtattions? But detail makes a difference in SEO Memations?

Slide 7

Slide 7 text

“It is easy to get good looking things from AI, but it is hard to get it exactly how you want it.” Bartłomiej Korpus CTO @Surfer

Slide 8

Slide 8 text

you need 3 areas covered! To get this

Slide 9

Slide 9 text

3 ingredients for growth STRATEGY DATA SCALE

Slide 10

Slide 10 text

1. Lack of strategy No clustering and research by search volume.

Slide 11

Slide 11 text

Why search volume? Traffic potential is a shiny object. People still believe it’s more important than providing comprehensive information around the niche using supporting pages.

Slide 12

Slide 12 text

What is clustering? Grouping keywords based on Google search results. If two keywords have the same results, they are a part of the same cluster. Target these keywords on the same URL.

Slide 13

Slide 13 text

Clustering tools

Slide 14

Slide 14 text

Surfer, scientiﬁc approach more than just comparing two SERPs

Slide 15

Slide 15 text

Compare all the keywords Find related topics Analyze semantics

Slide 16

Slide 16 text

Let’s see the clusters

Slide 17

Slide 17 text

This graph shows the relationship between keywords

Slide 18

Slide 18 text

This graph shows the semantic relationship between articles sorted by topic

Slide 19

Slide 19 text

All these calculations lead us to…

Slide 20

Slide 20 text

Semantics + Google data =

Slide 21

Slide 21 text

#rankboners This is how it should look like, after you focus on a niche instead of search volume.

Slide 22

Slide 22 text

Cover neighbourhoods not keywords Publish complete, interlinked silos.

Slide 23

Slide 23 text

2. Real time data Supposed limitation: GPT is trained on a dataset with a cutoff point.

Slide 24

Slide 24 text

2. Real time data Reality: It’s about data for OPTIMIZATION, not fresh concepts and news.

Slide 25

Slide 25 text

Google it!

Slide 26

Slide 26 text

Google it! ● AI can read ● AI can segment ● AI can find inspiration ● Just like your writers are doing research

Slide 27

Slide 27 text

Chat GPT + Web Pilot Single prompt article shortcut

Slide 28

Slide 28 text

Miedtattions? Remember this? Memations?

Slide 29

Slide 29 text

3 ingredients for optimization ENTITIES & TOPICS FACTS USER INTENT

Slide 30

Slide 30 text

Chat GPT + Web Pilot ● Research ● Segment ● Instruct ● Generate ● You can automate that process!

Slide 31

Slide 31 text

Surfer AI ● Analysis ● Optimization data ● Writing

Slide 32

Slide 32 text

Surfer AI ● Analysis ● Optimization data ● Writing

Slide 33

Slide 33 text

Surfer AI ● Analysis ● Optimization data ● Writing

Slide 34

Slide 34 text

400k tokens used 300k words long AI conversation to generate 3k words long article.

Slide 35

Slide 35 text

Surfer AI article published 6 months ago

Slide 36

Slide 36 text

Inspire AI with data Not raw articles but segmented bits of information.

Slide 37

Slide 37 text

3. Inferior AI engines GPT has many variants since 2018. Don’t fall in love with cheap.

Slide 38

Slide 38 text

Generative AI evolution

Slide 39

Slide 39 text

The best or? ● Small context window ● Repetitions ● No analytical skills ● Not following instructions ● Skewed facts, hallucinations

Slide 40

Slide 40 text

A lot of labour (that’s expensive)

Slide 41

Slide 41 text

Cost comparizon ● $25 in AI credits ● 15 minutes minor adjustments worth ? ● $25, 15 minutes - scalable AF ● $1 in AI credits ● 1h on editing ● 1h on optimization ● $1, 2 hours - not scalable Cheap long form AI Premium long form AI ● $1 in AI credits ● 2h on writing ● 1h on optimization ● $1, 3 hours - not scalable AI assisted writing

Slide 42

Slide 42 text

Calculate your margin Aim for the best quality and least time consuming option.

Slide 43

Slide 43 text

Case study

Slide 44

Slide 44 text

Team Tomasz Niezgoda Head of Marketing Bartek Barcik SEO Expert Zuzanna Roguska SEO Copywriter

Slide 45

Slide 45 text

Planned with clusters Human edited Jasper Content Surfer Optimized What did they do?

Slide 46

Slide 46 text

It works! Well planned and linked website thrive.

Slide 47

Slide 47 text

Penalized AI? Well planned and linked website receives a hit.

Slide 48

Slide 48 text

A minute of silence …

Slide 49

Slide 49 text

Trend continues Traffic and impressions going down.

Slide 50

Slide 50 text

Replacing the articles ● Batch by batch 30 pages at the time ● 150 pages replaced in total ● First test of Surfer AI

Slide 51

Slide 51 text

Single replaced article stats

Slide 52

Slide 52 text

Single replaced article stats

Slide 53

Slide 53 text

Single replaced article stats

Slide 54

Slide 54 text

Single replaced article stats

Slide 55

Slide 55 text

The overall results

Slide 56

Slide 56 text

It is 💩 content, not AI content a problem You can recover penalized AI with quality AI content.

Slide 57

Slide 57 text

Case Study 2 & 3 Dean Scaduto CEO @ Dino’s Digital

Slide 58

Slide 58 text

Rollercoaster ● EAT drop ● Scaled up content production ● Investigating the dip

Slide 59

Slide 59 text

Testing our approach Goal: Improve Topical Authority ● Find most relevant categories ● Create AI articles ● Publish complete, interlinked silos ● Analyze ranked keywords (too late)

Slide 60

Slide 60 text

YMYL site recovery after EAT issues

Slide 61

Slide 61 text

Biggest losers ● Benefits of not drinking for 3 and 6 months ● General keywords lost ● Ranked more than covered

Slide 62

Slide 62 text

SERP ● No time frame specific ● All general titles ● No doubt we lost

Slide 63

Slide 63 text

3rd Biggest loser ● Benefits of not smoking weed ● How to quit? We don’t mention. ● Ranked more than covered

Slide 64

Slide 64 text

SERP ● How to quit is a main keyword ● We didn’t even mention how to ● No doubt we lost

Slide 65

Slide 65 text

Today’s screenshot

Slide 66

Slide 66 text

What did we learn? ● Mass generation is just the start

Slide 67

Slide 67 text

What did we learn? ● Mass generation is just the start ● Dive into data

Slide 68

Slide 68 text

What did we learn? ● Mass generation is just the start ● Dive into data ● Be proactive, not reactive like us

Slide 69

Slide 69 text

12 months ﬂat ● Stabilized rankings ● Established authority

Slide 70

Slide 70 text

Flatline

Slide 71

Slide 71 text

Testing our approach Goal: Improve Topical Authority ● Find most relevant categories ● Create AI articles ● Publish complete, interlinked silos

Slide 72

Slide 72 text

Exacution

Slide 73

Slide 73 text

Explosive growth

Slide 74

Slide 74 text

Mission accomplished! Clusters + AI = 🔥

Slide 75

Slide 75 text

Entities & topics from SERP Inferior LLMs generate fast but take more time to edit Use best AI engines Analyze data Make your keyword research data driven Publish whole categories Key Takeaways

Slide 76

Slide 76 text

Thank You 5 FREE Surfer AI credits on top of your subscription. surferseo.com/seologer Find me here or reach me at: [email protected]