Slide 1

Slide 1 text

Why Google is taking aggressive action on SERP data collection and what to do about it Ray Grieselhuber @raygrieselhuber

Slide 2

Slide 2 text

No content

Slide 3

Slide 3 text

The rank trackers are dead!

Slide 4

Slide 4 text

Sound familiar?

Slide 5

Slide 5 text

Google makes some changes, everything is dead

Slide 6

Slide 6 text

I also said rank tracking is obsolete in 2023

Slide 7

Slide 7 text

Why would I say this?

Slide 8

Slide 8 text

Rank Tracking → SERP Analytics

Slide 9

Slide 9 text

Semantics?

Slide 10

Slide 10 text

How many reports still look like this?

Slide 11

Slide 11 text

SERPs have a shape and features that influence human behavior

Slide 12

Slide 12 text

(We’ll talk about agent behavior at the end)

Slide 13

Slide 13 text

SERP data belongs to all of us

Slide 14

Slide 14 text

If there is data ABOUT you or data that AFFECTS you…it is YOUR data

Slide 15

Slide 15 text

Paid for by ads targeted to us (using our data)

Slide 16

Slide 16 text

SERP data is the most valuable and underrated source of behavioral data on the planet

Slide 17

Slide 17 text

It is both a reflection and influencer of human behavior

Slide 18

Slide 18 text

Google is taking aggressive action to prevent the collection of data that belongs to us

Slide 19

Slide 19 text

A timeline:

Slide 20

Slide 20 text

Why are they being so aggressive?

Slide 21

Slide 21 text

The threat from AI companies

Slide 22

Slide 22 text

The index is the prize

Slide 23

Slide 23 text

The generated AI answers are also valuable

Slide 24

Slide 24 text

The agent-driven future is an existential threat to their current ad business

Slide 25

Slide 25 text

They don’t care about companies in our space

Slide 26

Slide 26 text

January 2025: SERPpocalypse

Slide 27

Slide 27 text

September 2025: Num=100 Removal

Slide 28

Slide 28 text

December 2025: SerpApi Lawsuit

Slide 29

Slide 29 text

May 2026: “Drop and Pops”

Slide 30

Slide 30 text

No content

Slide 31

Slide 31 text

July 2026: NetNut Takedown

Slide 32

Slide 32 text

August 2026: Goto Links

Slide 33

Slide 33 text

September 2026: The Banhammer, followed by an eerie silence…

Slide 34

Slide 34 text

October 2026: Poisoned SERPs

Slide 35

Slide 35 text

October 2026: Goto Links, Part Deux Olivier de Segonzac: https://www.linkedin.com/feed/update/urn:li:activity:7521955412418580481/

Slide 36

Slide 36 text

Other weirdness: Made up subdomains in search results

Slide 37

Slide 37 text

There are always countermeasures

Slide 38

Slide 38 text

Our response

Slide 39

Slide 39 text

Focus on top of the SERP (2-3 pages)

Slide 40

Slide 40 text

Real-time poisoning detection

Slide 41

Slide 41 text

Data labeling

Slide 42

Slide 42 text

Observability

Slide 43

Slide 43 text

Second-order observations on first-order data

Slide 44

Slide 44 text

The key question now is: How confident are you in this data?

Slide 45

Slide 45 text

A simple answer does not suffice

Slide 46

Slide 46 text

Accuracy is the wrong goal in longitudinal monitoring Mike King https://ipullrank.com/accuracy-vs-precision

Slide 47

Slide 47 text

You’re looking for patterns across a long range of samples

Slide 48

Slide 48 text

Removing deliberate sabotage such as poisoning is critical

Slide 49

Slide 49 text

The question is not: how confident are you in the data?

Slide 50

Slide 50 text

The question is: what does your data pipeline do in order capture samples reliably and eliminate and / or label anomalies?

Slide 51

Slide 51 text

Clearly label outages and errors

Slide 52

Slide 52 text

Send regular updates to stakeholders

Slide 53

Slide 53 text

Trusting the data is about trusting the process

Slide 54

Slide 54 text

Increased collaboration among providers is a positive outcome

Slide 55

Slide 55 text

The SERP of the (near) future

Slide 56

Slide 56 text

More generative features

Slide 57

Slide 57 text

Extensible https://blog.google/innovation-and-ai/technology/developers-tools/connected-apps-one-place/

Slide 58

Slide 58 text

Multi-turn

Slide 59

Slide 59 text

Very interactive

Slide 60

Slide 60 text

Maybe programmatic

Slide 61

Slide 61 text

The key question: What does Google’s ad model look like when the majority of SERP traffic is agentic?

Slide 62

Slide 62 text

It’s coming

Slide 63

Slide 63 text

The transition to local agent harnesses makes it much more difficult to block bots

Slide 64

Slide 64 text

These agents are acting on behalf of humans and sometimes using their hardware

Slide 65

Slide 65 text

Rankings don’t matter but SERP Features do

Slide 66

Slide 66 text

Ad-blindness is built in

Slide 67

Slide 67 text

Open source harnesses + open weight models are key

Slide 68

Slide 68 text

Stay on top of the releases in AI models, especially the open ones

Slide 69

Slide 69 text

Takeaways:

Slide 70

Slide 70 text

Join us at FOUND Tokyo 2027

Slide 71

Slide 71 text

Thank you! (+ link to slides) https://raygrieselhuber.com