Miranda Gahrmann - SEO in the era of AI - October 2025.pdf

Curious about how AI is shaking things up for SEO?

In this presentation, Miranda Gahrmann dives into the latest trends around AI chatbots like ChatGPT and Perplexity, showing how search is evolving and what that means for anyone working in SEO.

If you’ve been wondering how to keep your site visible when users are searching in totally new ways, don’t miss these practical tips on adapting fast, making the most of AI-powered tools, and staying ahead in a changing digital world

1. MIRANDA GAHRMANN
SEO IN THE
ERA OF AI
OCT 2025

2. MIRANDA GAHRMANN
•17 Years of experience in SEO

•Master in Digital Marketing
•Master of Brand Management
•> Master in Digital Media Design
ABOUT ME

4. CHANGES IN
THE SEARCH
LANDSCAPE

5. ●Since ChatGPT launched in Nov 2022, AI platforms have
quickly gained popularity

●Our main focus was to optimize for Google, but now our
scope has become broader
1. SEARCH BECOMES MORE FRAGMENTED

6. ●In terms of monthly visits, Google still has by far the
largest share

●Based on August data, Google had 67% of monthly
sessions, YouTube 24%, ChatGPT 5%, and Bing 3%
○Doesn't include local domain names or apps

1. SEARCH BECOMES MORE FRAGMENTED
Source: SimilarWeb

7. ●In terms of traffic, Google is by far the largest driver

●Bing and DuckDuckGo drive more traffic than ChatGPT

●This Ahrefs dashboard shows SEO vs. AI traffic across
60,000+ websites
1. SEARCH BECOMES MORE FRAGMENTED

8. ●Statcounter tracks the market share of AI chatbots

●The main players are:
○ChatGPT: 81.1%
○Perplexity: 10.8%
○Copilot: 4.1%
1. SEARCH BECOMES MORE FRAGMENTED

9. ●Search engine usage is declining among younger users

●Gen Z (ages 13–28) uses Google for searches 25% less
than Gen X (ages 45–60)

●40% of Gen Z prefer TikTok or Instagram over Google
2. GEN Z PREFERS TIKTOK OVER GOOGLE

10. ●SimilarWeb data shows that visits to Google are declining

●Google claims that the number of queries is increasing

●Both could be true, as AI overviews and AI Mode could
lead to more queries per session
3. MORE QUERIES PER USER

11. ●AI chatbots change how people search online
○More specific and context-rich

●Prompt frameworks help to get better results from AI

●Users can go back to previous conversations
4. QUERIES ARE CHANGING
GOOGLE’S
SUGGESTED PROMPT
FRAMEWORK

●TASK
○PERSONA
○FORMAT

●CONTEXT

●REFERENCES

●EVALUATE

●ITERATE

12. Specific and
context-ric
h prompt

13. Traditional
search vs AI
mode
AI Overview
yes, regular no

14. Bing screw
up…

15. ●Thanks to AI tools, the web will be flooded with
low-quality pages

●Search engines will raise the bar and further prioritize
original, high-quality content
5. MASS CONTENT PRODUCTION

16. ●Google shows info directly from websites

●AI tools generate answers by mixing sources, and
hallucinate…
6. INCREASED NEED FOR FACT-CHECKING

17. ●Google allows the use of AI for content creation

●But…. final output needs to be original and high-quality
6. INCREASED NEED FOR FACT-CHECKING

18. ●In May 2024, AI Overviews went live in the US
●In Oct 2024, Google rolled it out globally

●The impact:
○Clicks dropped significantly (34% - 46%)
○Gap between impressions and clicks widened
7. INCREASE IN ZERO-CLICK
RESULTS

19. ●In May 2025, AI mode rolled out in the US
●On Oct 7th, AI Mode rolled out in NL & BE

●Links are often to URL fragments (#)

●So far, AI Mode isn't the default view

8. GOOGLE HAS ROLLED OUT AI
MODE

20. ●On mobile, the link cards are only present at
the very end of the answer

●Tapping sites or any anchor brings up an
overlay with the links
8. GOOGLE HAS ROLLED OUT AI
MODE

21. ●In May 2025, Gemini was integrated into
Chrome in the US

●In July 2025, Perplexity launched its own
browser Comet, with an integrated AI assistant
9. AI ASSISTANTS IN BROWSERS

22. ●AI agents are more autonomous than AI assistants
○Assistants are reactive and perform tasks when asked
○Agents are proactive and work independently

●In Jan 2025, OpenAI launched Operator
●In March 2025, Manus went live
10. GROWING ADOPTION OF AI AGENTS

23. ●AI assistants search, synthesize, and summarize info
from the web to give an answer

●Users may never see the original source

●We need to optimize for both users and bots
11. OPTIMIZING FOR BOTS AND USERS

24. ●Google ranks and shows entire pages

●AI chatbots extract passages from multiple sources

●Sources are less visible compared to Google SERPs
12. PAGES VS PASSAGES

25. ●In Nov 2024, Perplexity launched a direct shopping
integration

●ChatGPT will roll this out soon
13. DIRECT SHOPPING INTEGRATION

26. Shopping
overlay

27. Finalize
within
Perplexity

28. ●In May 2025, Google rolled out Shopping with AI Mode
in the US

●You can track products and get alerts when prices drop
13. DIRECT SHOPPING INTEGRATION

29. Shopping
with AI
Mode

30. ● From the alert, you can buy the product directly
13. DIRECT SHOPPING INTEGRATION

31. ●On Oct 6, ChatGPT launched apps within ChatGPT

●This allows companies like Booking.com and Expedia to
display their listings directly in ChatGPT

●Integration through the Model Context Protocol (MCP)
13. DIRECT SHOPPING INTEGRATION

32. ●Currently no tool available that provides accurate prompt
research data

●Emerging tools make predictions using existing data
sources
14. AVAILABILITY OF SEARCH DATA

33. ●Google Search Console data includes AI queries
●Can be identified by unnatural patterns and zero clicks
●Common phrases include:
○Evaluate
○Can you list
○Reply in English
15. LESS RELIABLE REPORTING DATA

34. ●Scrapers & agents can inflate the ChatGPT-User hit count

●ChatGPT-User refers to user actions and Custom GPTs

●According to the data, ChatGPT-User hits make up 28% of
Googlebot's traffic, which is unlikely

15. LESS RELIABLE REPORTING DATA

35. ●More tools offer prompt tracking, but it’s often unclear
which models they use or whether RAG is included

●By default, the API doesn’t include Retrieval-Augmented
Generation (RAG)
○Perplexity’s API does include web search

15. LESS RELIABLE REPORTING DATA

36. Same time,
different
answers…

37. ●Around Sep 11, Google removed the ability to extract 100
results at once
○https://www.google.com/search?q=most+scenic+marathons+in+the+world&num=100

●This change has affected SEO ranking trackers and
Google Search Console data
○No actual traffic has been affected
15. LESS RELIABLE REPORTING DATA

38. WHAT’S NEXT?

39. ●Web crawling is a primary source of training data for
Large Language Models

●If your site doesn’t cover topics you want to be known
for, high chance LLMs won’t include you in the
conversation
WE STILL NEED WEBSITES

40. Not all
prompts
are
searches

OpenAI’s Sep 2025
research is based on 1.5
million conversations

41. ●The conversion rate from ChatGPT traffic is higher than
SEO traffic

●Users get educated by LLMs before they reach the site

●For a client I work with, the CR is about 3x higher
CONVERSION RATE IS HIGHER

42. ●Product traffic seems quite stable
○People still want to buy a product or apply for a job if
that’s their intention

●Informational traffic has been impacted, primarily by AI
Overviews
PRODUCT TRAFFIC IS QUITE STABLE

43. HOW TO APPEAR IN
GOOGLE
AND LLMS?

44. ●Make sure robots.txt doesn’t block crawling

●Check log files to see if any bots are experiencing issues

●Use IndexNow to notify Bing of new or updated content

1.ARE PAGES ARE CRAWLABLE AND
INDEXABLE?

45. ●Create original, in-depth pages for key topics

●Highlight each topic from multiple angles

●Include original data, statistics, or unique insights
2. CREATE IN-DEPTH PAGES

46. ●Break text into logical blocks, focused on a single idea

●Write in a clear, conversational tone
○Use direct language when presenting factual claims

●Add a summary like Key Takeaways
3. STRUCTURE CONTENT FOR EXTRACTION
Duane Forrester- June 2025

47. ●Where appropriate, include:
○Question-based headings
○Bullet points
○Tables
○FAQ section

3. STRUCTURE CONTENT FOR EXTRACTION

48. Stay within
Google’s
guidelines
●“Google's ranking systems aim to reward original, high-quality content
that demonstrates qualities of what we call E-E-A-T: expertise,
experience, authoritativeness, and trustworthiness.”

●“Using automation—including AI—to generate content with the primary
purpose of manipulating ranking in search results is a violation of our
spam policies.”

●“Multiple systems work in various ways to show the most reliable
information possible, such as to help surface more authoritative pages
and demote low-quality content and to elevate quality journalism”

49. ●Fresh articles seem to perform better
○Regularly update existing content and add new
information

●Use IndexNow or resubmit pages in Google Search
Console to help search engines notice changes faster
4. OPTIMIZE FOR FRESHNESS

50. ●Analyze your content the way AI does

●Use tools to review embeddings and measure semantic
similarity

●Compare similarity scores with your competitors
5. REVIEW PAGES LIKE GOOGLE AND AI

51. ●Add schema markup to relevant pages

●Microsoft has confirmed schema helps with Copilot

●Other LLMs as ChatGPT often don’t see the structured
data by design as it’s stripped during preprocessing
6. IMPLEMENT STRUCTURED DATA

52. ●Page speed is a ranking factor for Google and improves
user experience
○This is more than just meeting Core Web Vitals

●Faster pages allow more URLs to be crawled within the
same crawl budget
7. HAVE FAST, MOBILE-FRIENDLY PAGES

53. ●LLMs are becoming multimodal, able to process and
generate text, images, audio, and video

●Help search engines and LLMs understand visual content
○Provide transcripts for videos and audio
○Add descriptive alt text to all images
8. LEVERAGE MULTIMEDIA

54. ●Increase your brand’s presence on reputable websites

●Get mentioned in news articles, industry blogs, reviews,
and partner sites

●Earn mentions through digital PR and thought
leadership
9. BE PRESENT ON OTHER PLATFORMS

55. ●Review sites like Trustpilot and ConsumerAffairs are
used as input for LLMs

●Monitor your brand online, step in where needed, and
actively work to generate positive reviews
10. IMPROVE YOUR BRAND’S REPUTATION

56. ACTIONS
SPECIFICALLY FOR
LLMS

57. ●Sign up for the Merchant Program from both ChatGPT and
Perplexity

●Create apps for ChatGPT so users can ask your company
directly about your listings
1.BE PART OF THE ECOSYSTEM

58. ●Check the log files for 403 status codes or other signs
that AI bots are being blocked
○Especially if you use Cloudflare
2. REVIEW LOGS DATA

59. ●Prompt tracking has its flaws, but provides a useful
indication of your visibility vs competitors

●Use tools like Rankshift, Profound, Peec AI, and Otterly

●Set up your own tracking in Google Sheets & Apps Script
3. MONITOR OUR AI VISIBILITY

60. ●Note which sources are cited by LLMs

●Identify high-value sites where your competitors are
mentioned, but your brand is not
○For targeted outreach
4. ANALYZE CITATIONS IN AI TOOLS

61. ●Being part of authentic, ongoing conversations on
platforms like Reddit
○Also helps with brand’s visibility, trust, and authority

●LLMs and search engines that increasingly reference
these discussions
5. BE INCLUDED IN ONLINE CONVERSATIONS

62. ●Articles that include comparisons are often used as
prominent sources in AI tools
6. GET FEATURED IN COMPARISON ARTICLES

63. ●An llms.txt file provides a summary of the site in markdown
○https://www.indeed.com/llms.txt

●It's a proposed standard but hasn't been recognized yet

●It’s quick to generate and doesn’t cause harm
7. CONSIDER AN LLMS.TXT

64. HANDY TOOLS

65. CHROME
EXTENSIONS
●SEO SIMILARITY CHECKER

●AI CITATION ANALYSIS

●AI SEARCH IMPACT ANALYSIS

66. THE END
QUESTIONS?


Miranda Gahrmann
●linkedin.com/in/mirandagahrmann

●https://bit.ly/seo-benelux-ai

67. EXTRA :)

68. COMMON
MISCONCEPTIONS

69. ●Misconception: Many believe AI is a brand-new disruptor
in SEO
●Reality: AI has been part of search engine algorithms
since the 2013 Hummingbird update. What changed with
the launch of ChatGPT in Nov 2022 is that advanced AI
became directly accessible to the public
“AI IS NEW IN SEO”

70. ●Misconception: Some see AEO as a totally separate or
brand-new discipline
●Reality: AEO as a term is new, but the underlying strategy
of optimizing for direct answers has been a part of SEO
since at least 2012 with the introduction of Google’s
Knowledge Graph and featured snippets
“ANSWER ENGINE OPTIMIZATION (AEO) IS
NEW”

71. ●Misconception: People have been saying “SEO is dead”
since 1997
●Reality: Every time Google or other search engines
updates their algorithms, or new technology emerges,
this myth resurfaces
●SEO continues to be essential, but the tactics have shifted
“SEO IS DEAD”

72. EVOLUTION OF AI IN
GOOGLE

73. FOCUS
Keyword matching, on-page signals, and early link
analysis
MAJOR UPDATES
●Panda (2011): Penalized low-quality, thin, or
duplicate content
●Penguin (2012): Targeted manipulative link
schemes and web spam
KEYWORD ERA
(1998–2013)

WHAT IT MEANT FOR SEO
●SEO was about keyword usage
●Panda and Penguin forced a shift toward quality
content and ethical link building

74. FOCUS
Understanding intent and meaning behind queries
MAJOR UPDATES
●Hummingbird (2013): Improved Google’s ability to
understand more complex queries, focused on intent
●RankBrain (2015): Introduced machine learning to
interpret search intent, context, and user behavior
●BERT (2019): Improved Google’s understanding of
natural language and context within queries
INTENT ERA
(2013–2019)
WHAT IT MEANT FOR SEO
●Shifted to focus on search intent, semantic
relevance, and user experience

75. FOCUS
Multimodal understanding, context, and generative AI
MAJOR UPDATES
●MUM (2021): Google’s Multitask Unified Model
understands information across text, images, and
more, answering complex, multi-part questions
●AI Overviews (2024): Answers directly in the SERPs
CONTEXT ERA
(2020–PRESENT)

WHAT IT MEANT FOR SEO
●Optimizing for both traditional search and a
growing range of AI-driven platforms
●Making content easily extractable and useful for
answer engines and generative AI tools

76. WHERE DO LLMS
GET THEIR DATA
FROM?

77. ●Web crawling is a primary source of pretraining data for
Large Language Models
●Pretraining datasets are typically built using large-scale
web crawls, such as Common Crawl
●Most of the crawled web pages are excluded from
pretraining due to low data quality
WHERE DO LLMS GET THEIR DATA FROM?

78. ●Web pages from large-scale crawls like Common Crawl
●Wikipedia and digitized books
●Community platforms like Reddit and Stack Exchange
●Academic papers and News articles
●Code repositories like GitHub and Kaggle
●Video subtitles and transcripts from platforms like
YouTube
PRIMARY DATA SOURCES FOR LLMS

79. ●Perplexity AI augments their responses with web search,
to provide up-to-date information
●Perplexity uses its own search index, so it’s not
performing a real-time web search
●The underlying LLMs are not updated or learning in real
time
STATIC VS. REAL-TIME KNOWLEDGE

80. ●RAG helps LLMS give better answers by pulling in info
from external sources
●Retrieval: When a user submits a query, the system
searches external knowledge sources
●Augmentation: Retrieved info is added to original prompt
●Generation: The LLM uses both its internal knowledge
and the added information to generate a response
RETRIEVAL-AUGMENTED GENERATION (RAG)

81. ●Grounding is the process of anchoring an AI system’s
outputs to verifiable, real-world data or sources
●It ensures that the model’s responses are based on
specific, factual information
THE ROLE OF GROUNDING IN RAG

82. INTERNAL
COMMUNICATION

83. ●Your manager sees the AI developments and ads too
●Be proactive in updating them on how you're using AI to
improve efficiency while maintaining high quality
●Don’t copy paste AI-generated advice
○if you do, why would they need to keep you?
●Use AI thoughtfully and think critically
INTERNAL COMMUNICATION

84. SO, IS THIS STILL
SEO?

85. ●A lot of acronyms have popped up lately, each targeting
different platforms or aspects of search behavior
●We’re keeping it simple:
○We target search behavior, wherever the user is
●And we’ve got the perfect acronym for that...
ACRONYMS

86. SEO =
Search Everywhere Optimization

RAND FISHKIN

87. THE END
QUESTIONS?


Miranda Gahrmann
●linkedin.com/in/mirandagahrmann

●https://bit.ly/seo-benelux-ai