How to Manipulate AI Search (Before It Manipulates You)

Overview

This episode discusses how AI-driven search systems (e.g., Google AI Overviews, ChatGPT, Claude, Perplexity, Grok) decide what to recommend and cite—and how marketers can “optimize for” those systems.

It blends traditional SEO tactics with newer approaches focused on distribution, reputation, and proof.

Key points and arguments

1) YouTube is a major (underused) driver for AI citations

• The guest argues that YouTube is heavily leveraged inside Google’s AI Overviews, citing an external claim that YouTube is among the most cited sources (and/or most cited/clicked sites) in Google’s ecosystem.
• Many SEOs, they suggest, underutilize YouTube for building “AI visibility.”
• YouTube content can support both conventional rankings and AI-overview citations.

2) Brand reputation can directly affect AI recommendations—and paid performance

• Example: a known brand was heavily hammered on Trustpilot.
• The episode claims this negative sentiment showed up in AI overview language (“don’t choose this brand”) and also increased paid lead costs sharply (from ~£126 to ~£442 per lead).
• The proposed fix wasn’t simply “fix Trustpilot” first; it was shifting where reviews are surfaced/circulated so the AI sees a different consensus.

3) “Compact keywords” and purchase-intent SEO funnels (ads/links discussed as conversion assets)

• The episode promotes “compact keywords”:
  • Instead of long “answer” content, publish many small pages targeting buyer intent.
  • These pages are designed to rank and convert.
• Typical landing page length mentioned: ~415 words per page.
• The framing: it generates measurable leads, “works with AI,” and can require less effort than traditional SEO.

4) How AI could choose between “good SEO/bad reviews” vs “ok SEO/no reviews”

• The guest argues AI won’t just choose the option with the best raw SEO.
• It may “fan out” and run deeper checks, including:
  • founder/background signals
  • additional supporting signals
  • broader performance/quality metrics
• Takeaway: when AI compares options, you need both:
  • discoverability (SEO)
  • trust/reputation signals (reviews, testimonials, proof)

5) Reviews strategy: “swap out” review platforms and amplify consensus

A playbook for improving AI outcomes using reputation signals:

• If Trustpilot becomes too negative (“burnt to the ground”), shift review volume to another platform (examples mentioned: FIFO, Reviews.io, etc.).
• Mechanism described: review sites can rise in AI/SEO visibility because they keep receiving traffic and referral traction (e.g., post-service follow-up emails sending customers to a specific review site).
• “Proof layers” approach:
  • Use multiple formats (video, images, text reviews).
  • Include full transcripts on the review page (while trimming on sales pages for UX).
  • Potentially create additional “executive summary” pages using AI-generated summaries of reviews-site content.

6) Retargeting as an “easy” conversion booster (and part of the funnel)

• Visitors to review pages who don’t convert can be retargeted on Meta/Facebook/Instagram.
• The claim: this is one of the cheapest ways to extract more value from warm traffic that already showed intent.

7) Customer service as an SEO/AI-reputation lever

• Poor or inconsistent customer service can generate public complaints.
• Those complaints then show up on review sites and social platforms, harming AI/AI-overview outcomes.
• The episode argues that even small improvements in responsiveness can prevent negative reviews.
• Strong service can allegedly create “parasite SEO” effects, where customers generate content and recommendations.

8) Gray/black hat niche tactics mentioned: manipulating clicks/traffic

• The guest discusses manipulating click/traffic patterns to influence rankings.
• These tactics are described as expensive and requiring infrastructure (e.g., IP management).
• They contrast this with “white hat” virality, where organic attention spikes can be interpreted positively by ranking systems.

9) Testing and AI mechanics: different LLMs crawl/cite differently

The guest describes experimentation and hypotheses about how citations work:

• Grok: appears to drop citations quickly after page deletions, suggesting limited cache.
• ChatGPT/Claude: citations may persist longer, implying a preliminary cache before deeper crawling/common crawl.
• ChatGPT is described as having multiple “bot” stages, including:
  • common crawl
  • a page-visiting bot
  • a bot that looks at “10 blue links” metadata
• A citation-overlap tracking sheet suggests:
  • significant overlap between some systems (e.g., Google AI ↔ Perplexity; Perplexity ↔ Claude; Perplexity ↔ ChatGPT)
  • differences attributed to underlying search sources and crawling patterns

10) Practical recommendations for getting AI “recommendations”

For established brands:

• Emphasize review amplification and (when appropriate) comparisons.
• Potentially use YouTube heavily, including hiring a dedicated YouTube person.
• Use “kitchen sink” coverage when starting from zero:
  • reviews
  • pros/cons comparisons
  • video/testimonials
  • social proof
  • broad distribution

For less-established/local brands:

• Focus on fundamentals and buyer intent aligned with query “fan-out.”
• Target structured local intent, not only generic queries (example given: hedge trimming/pricing modifiers rather than only “tree surgeon”).
• Reinforce that AI still does due diligence using reviews and proof.

11) Backlinks, schema, and listicles: revising “SEO myths”

• Listicles
  • Defended strongly.
  • The claim: AI citations in the guest’s sample were often listicles.
  • They can work if handled transparently (not deceptively).
• Schema
  • Framed as less of a direct AI-overview ranking factor than many believe.
  • More about “tying loose ends,” enabling knowledge panels, and indirectly improving trust/consistency.
• Backlinks
  • Claimed to be less about raw link equity/PageRank.
  • More about “theming” (who you are and who you serve).
  • Mentions are framed as increasingly important.

12) Affiliates and publishers: pivot away from “nameless SEO sites”

• Affiliate strategies are portrayed as shifting:
  • brands prefer partnerships with recognizable “personality/brand” accounts rather than anonymous review pages.
• Publishers are advised to:
  • sell their own product, or
  • build a branded SaaS/asset
• Rationale: AI may strip affiliate links even if it still recommends the underlying product.

Final advice themes (recurring throughout)

• Build consensus signals AI can aggregate: reviews + testimonials + multi-format proof + distribution (especially YouTube and social).
• Use claim → frame → prove:
  • make a strong claim
  • support it with external proof
  • ensure multiple independent sources repeat it
• Experiment and iterate:
  • SEO changes are reversible compared to fear-based inertia.
• Treat customer service as foundational:
  • it drives reputation signals, which then influence AI summaries and recommendations.

Presenters / contributors

• Edward Storm (podcast host)
• Casar Dash / Caser Dash (guest; marketing/SEO specialist; referenced as the “legend”)