When MCP Servers are (and aren’t) the Right Choice

Transcript

1. When MCP Servers are (and aren’t) the Right Choice Gus Pelogia Sr. SEO Product Manager Indeed guspelogia.com
2. LLMs, 2024 Before MCP Keyword Search Volume (US) used cars 800k used cars near me 420k used cars for sale 125k
3. LLMs, 2026 Before MCP Keyword Search Volume Hallucinated Search Volume Real* used cars 800k 673k used cars near me 420k 450k used cars for sale 125k 301k *Matching SEO tool of choice
4. Model Context Protocol An open-source standard that enables AI models to securely connect with external tools In other words: APIs inside LLMs
6. Why MCPs are so powerful
7. Why MCPs are so powerful LLM MCP Find the top 5 "seo conference" related keyword by search volume Calls Sistrix/SE Ranking/Ahrefs keyword database to reply
8. Why MCPs are so powerful LLM MCP 1 MCP 2 Find the top 5 "seo conference" related keyword by search volume Calls Sistrix/SE Ranking/Ahrefs keyword database to reply Use DataForSEO and return the SERP results for US, UK and Ireland
9. SEO MCP Servers Timeline Sistrix SE Ranking Ahrefs AlsoAsked Aug '25 Oct '25 Nov '25 Jun '26 DataForSEO Semrush Profound Screaming Frog Sitebulb May '25 Sep '25 Oct '25 May '26 Sep '26
10. Connect MCP to Claude Code
11. Keyword Research Ahrefs, Sistrix, Semrush Ps 1 : Same tasks can be performed by different tools Crawling Screaming Frog, Sitebulb SERP Scraping DataForSEO, SE Ranking
12. 1 Compare your favorite tools
13. Compare equivalent data points from different tools in one go Collect AI citations from AIO, AI Mode and ChatGPT Put them side by side
14. Citations / LLM Sistrix Ahrefs Google AI Overviews 52,202 18,427 Google AI Mode 17,575 22,697 ChatGPT 3,023 13,045 AI Citations — Total 72,800 54,169
15. How about a crawl comparison?
16. Standard configuration diff
17. 2 From PAAs to prompts
18. Check Semrush MCP and return the keywords [domain] ranks for in the top 3 organic results Run the first ten terms by AlsoAsked MCP and return me the questions listed Now use SE Ranking MCP to add the first level questions into prompt tracking for AI Mode
19. Claude Code SE Ranking Regular UI
20. Claude Code Profound UI
21. Tool Shout-Out: Sistrix
22. Sometimes, MCP is not the right tool When you need to identify RAG dependency
23. Prompt webSearchQueries What foods do you eat in a keto diet? What is magnesium critrate used for? What does turmeric does for the body? Can I influence an answer?
24. Gemini Grounding API
25. You can still request your tracked prompts from another MCP Gemini has an MCP, but Claude created a script instead webSearchQueries is the API field used. Not a probability score, only the result for this prompt, checked once Gemini Grounding API
26. 3 Use tracked keywords for live rankings on Google Maps
27. Check your regular SERP keywords on GMaps, by city Select relevant tracked keywords / prompts. Ask DataForSEO to scrape terms in list of locations Summarize most popular brands in the [top 5] spots
29. Sometimes, MCP is not the right tool When you need to scrape +20 SERPs
30. Live SERP Scraping at Scale
31. Use LLM to write a Pyhthon code to run as a skill or Google Colab It works, but it's slow. More than 20 SERPs, not worth waiting Claude will write Python script to run it - and do a few at a time Live SERP Scraping at Scale
32. 4 Analyse crawl for "striking distance” keywords
33. First, store the hints found on Sitebulb MCP for [domain] Then, use Semrush MCP to find keywords in positions 3-20 Display Sitebulb hints only for pages in this group
35. Sitebulb: Read Only + All Plans Included
37. 5 Compare similarity between AI Overview and organic results
38. Use DataForSEO to scrape AI Overview + SERP results. Saves as an .md file Store Gemini API key in .zshrc, so Claude can’t see your key Compare cosine similarity between AI Overview and selected pages
39. AIO vs Trad. Organic AIO vs AIO Sources keto diet keto diet Rank URL Similarity Rank URL Similarity 6 theguardian.com 0.8357 1 newsroom.heart.org 0.8286 4 newsroom.heart.org 0.8286 2 hopkinsmedicine.org 0.8271 2 hopkinsmedicine.org 0.8271 3 health.harvard.edu 0.7871 8 racgp.org.au 0.8268 4 healthline.com 0.7849 7 florre.co 0.8243 5 bhf.org.uk 0.7738
40. Is an MCP the right answer? Claude Colab Use MCP to get embeddings Run crawl on Screaming Frog with embeddings Create a skill to run on Claude Code Use Google Colab to get cosine similarity Next step: do I have pages with high similarity on my website?
41. 6 Perform a content gap analysis using rankings and published URLs
42. Ask Claude to crawl [Domain A] and [Domain B] using Screaming Frog Overlay the [top 3] ranking keywords for each domain Find (1) pages I already have, but don't rank and (2) pages I don't have at all.
43. These are local Irish brands in my city - I'm not affiliated with them
44. Slides: guspelogia.com/bseo-oct26