Video summary

The Future of SEO: Lily Ray on Google Updates, AI Search & GEO Spam

Main summary

Key takeaways

Educational

Main ideas / concepts covered

  • Google’s long-term pattern in core updates

    • Across many core updates (analyzed for ~10 years), Lily Ray says Google’s consistent goals are:
      • Improve relevancy and quality of results
      • Reduce spam and overly optimized SEO
    • During specific periods (e.g., COVID), Google has also prioritized high-authority sources (like government sites).

  • What “too much SEO” means (and why it’s counterintuitive)

    • “Too much SEO” is framed as tactics that:
      • Work initially by tricking algorithms
      • Become widely adopted (because they’re popularized in the industry)
      • Then get targeted by Google because the result set becomes low-quality/homogeneous
    • Common patterns she cites:
      • Scaling content via tricks
      • Creating tens of thousands of pages using the same pattern
      • Content that doesn’t add meaningful value (especially when it’s “SEO-first” rather than human-first)

  • Programmatic SEO vs spam

    • Programmatic SEO is described as often risky, though it can be legitimate when:
      • It’s thoughtful
      • It includes original data/insights Google wouldn’t otherwise have
    • Example of a high-risk style:
      • Pages generated around things like area codes, largely to serve ads—Google “doesn’t need 50,000 pages” about it.
    • Key distinction emphasized:
      • Unique + thoughtful + original value is safer than content that merely repackages what already exists.

  • GEO (Generative Engine Optimization) spam disasters

    • Ray argues GEO is still early, so some manipulative tactics can temporarily work—similar to earlier eras of SEO (e.g., before major anti-spam updates like Penguin / helpful content enforcement).
    • She expects countermeasures as systems learn what’s being manipulated.

  • Self-promotional listicles (GEO tactic) and potential backlash

    • Defined as articles like:
      • “Best X for Y” where the publisher ranks itself #1
    • Discussed as:
      • Something that may work for LLM citations briefly
      • Potentially becoming worse for brands if LLMs start detecting “suspicious” self-placement
    • She claims she has directly seen LLM behavior reflecting category-spam awareness (e.g., Claude adding disclaimers about spammy categories and preferring third parties).

  • Why “site:”/domain filtering may reduce spam in AI answers

    • She hypothesizes that AI models may use whitelisting-like domain sets for different query types, e.g.:
      • “Trustpilot / G2” for reviews
      • “Tripadvisor / Expedia / Booking / Yelp” for travel
    • Broader point: LLMs don’t need to treat all sites equally—spam and manipulation can drive domain-level trust filtering.

  • E-E-A-T as the sustainable path (human, not tool-first)

    • She repeats that a sustainable GEO/SEO approach is:
      • Become “a human being” known in the industry
      • Build a footprint through real people, original ideas, and authentic interaction
    • Suggested approach:
      • Use AI to assist with writing, but don’t let AI “do the thinking” entirely
      • Build presence across platforms where citations/trust may come from (social + community + user-generated ecosystems)

  • How Google judges E-E-A-T

    • She frames it as a very mixed system rather than one simple ranking factor:
      • Knowledge Graph entity associations
      • SERP features highlighting experts/contributors
      • Long-term references to E-E-A-T across quality guidance (but indirectly)
    • Overall: E-E-A-T is described as not easily manipulable; it’s more like “signals/vibes” derived from ecosystem patterns.

Practical guidance / methodology (detailed bullets)

A) What to do after losing 70–80%+ organic traffic (or during major drops)

  • First, rule out technical crawl/index issues
    • Accidentally noindexing content
    • robots.txt blocking
    • JS rendering problems (search engines can’t see content)
    • (If found, these are sometimes fast to fix)
  • If not technical, reverse-engineer likely causes
    • She says it almost always connects to shady SEO history (sometimes from 1–2 years prior)
    • Investigate:
      • Past agencies/consultants and what they changed
      • Patterns visible during a site crawl
      • “Problem elements” (she uses metaphors like “cancer”/problem sections)
  • Work with someone specialized in core-update patterns
    • Not all SEOs have update-recovery expertise.
    • She names examples of people known for algorithm-update pattern work (see “Speakers/sources” section).

B) “Offensive SEO” after fundamentals are fixed

  • If “defensive” basics (fixing known mistakes) are done, she suggests “offensive” should still be rooted in:
    • Content + social strategy rather than risky loopholes
  • Example offensive approaches she gives:
    • Original content based on survey data / original research
    • Content aligned with what the audience wants (not just what ranks)
    • Aim for:
      • Press/news mentions
      • Strong links earned naturally
      • Social discussion
    • Use formats that can also drive Google Discover visibility.

C) E-E-A-T / content trust-building checklist concept (implied from her guidance)

  • Build signals that your brand/experts are real and known:
    • Contributors/experts with recognizable presence
    • Original ideas and original content
    • Social profiles + engagement
    • Authentic publishing across relevant platforms
  • Avoid shortcuts that try to “fake” author expertise:
    • Examples include AI-generated “author images” (can work briefly, but is not the core of E-E-A-T)

D) Link earning approach she favors (safer “earned” strategy)

  • Prefer “linkable assets” that are hard to fake:
    • Surveys, original research
    • Original images/videos
    • Other content that takes significant effort and naturally attracts citations
  • Use outreach to earn links to assets you built, not to manipulate results
  • She strongly avoids positioning paid/dodgy link building as a recommended baseline, especially for long-term client safety.

E) What AI should/shouldn’t automate in SEO

  • Automate (generally acceptable):
    • Keyword/topic research
    • Reporting
    • Data analysis
  • Avoid automating “the publish button” at scale
    • Example risk: auto-updating thousands of articles based on AI-generated “freshness”
    • She warns this evolves from earlier spam techniques (“artificial refreshing”) into AI-based versions of spam
  • If updating content:
    • It must include unique/original/meaningful human oversight
    • Google can detect AI content patterns that aren’t backed by real editorial effort.

F) Content tactics: balances and warnings

  • Titles/keywords still matter, but modern SEO should emphasize:
    • Clear, human-friendly content that satisfies intent
    • Don’t just stuff keywords; Google understands semantics better now
  • She’s skeptical of overly technical “chunk answer” / embeddings tactics if everyone does them, because:
    • It could skew away from good human content.

“Helpful content update” takeaways (what she believes applies in 2026)

  • She says the update:
    • Broke many sites severely and some didn’t fully recover
    • Was likely over-aggressive in outcome relative to merit for many sites
  • Main lesson:
    • Even tactics that work (for months) can become liabilities once they’re widely copied and users complain.
    • Google then counteracts with stronger measures than expected.
  • Examples of tactics becoming popular and therefore riskier:
    • Scaled AI content with clicky formats
    • TLDRs at the top
      • She says TLDR can be helpful and not inherently “spam”
      • But if the industry homogenizes around SEO patterns, Google may demote what looks like manipulation
    • Tables of contents (TOC)
      • She previously used them and still believes they’re often good for users
      • Even if TOCs are “good,” she notes Danny Sullivan later criticized TOCs as a repeated SEO pattern; she treats TOC as “one clue among many,” not automatically harmful.

What distinguishes sites that survive updates vs collapse (her thesis)

  • Don’t assume something that worked for a long time will keep working.
  • She argues a core motivation of core updates is:
    • Demote SEO tactics that are working too well and are being widely emulated.
  • Strategic framing she offers:
    • The best outcome is to look more like a non-SEO-optimized entity:
      • Examples she gives include government institutions or trusted organizations
    • A “defensive SEO” mindset:
      • Address technical UX/conversion needs
      • But don’t exploit SEO tactics primarily to inflate traffic volume

SEO tactics that “still work” (and why people don’t discuss them much)

  • Internal linking
  • Conversion rate optimization
  • Adding trust/credibility information
    • More author/brand info and evidence of expertise “can’t hurt” in her view
  • She ties this to user behavior signals:
    • If content satisfies intent, users don’t pogo-stick (bounce back quickly), which supports rankings.

Affiliate websites (viability and constraints)

  • She says affiliate SEO remains hard and Google has cracked down, especially on low-quality product review strategies.
  • She argues it can still work long-term if:
    • You follow review guidelines
    • You provide real evidence you tested products (photos/videos, genuine trial)
  • Most people won’t do the level of proof required.

Speakers / sources featured (as named in the subtitles)

Primary speaker

  • Lily Ray (SEO/GEO expert; guest)

Interview host / podcast host (implied by “Edward Show”)

  • Edward (podcast host; name not explicitly provided in subtitles, but referenced as “Edward Show”)

Other people / organizations referenced

  • Google (Search Quality Guidelines; helpful content update; E-E-A-T mentions)
  • Danny Sullivan (Google; discussed post-helpful-content patterns)
  • Matt Cutts (historical Google spokesperson referenced)
  • Rand Fishkin (referenced via a prior episode)
  • Aleyda Solis (referenced for content on what to do after drops)
  • Glenn Gabe (Ahrefs/industry update recovery guidance referenced)
  • Marie Haynes (core update specialists referenced)
  • Christopher Long / Chris Long (mentioned in the discussion about screenshots; first name appears as Chris/Christopher)
  • Ahrefs (Glenn Gabe works/brand referenced; and tools/MCP referenced)
  • Anthropic (Claude referenced; companies “opening eyes”)
  • Claude (Anthropic model referenced repeatedly)
  • ChatGPT / OpenAI (referenced)
  • Google Search Console / Google Analytics / Google Discover (Google products referenced)
  • Trustpilot / G2 (example review platforms)
  • Tripadvisor / Expedia / Booking / Yelp (example travel platforms)
  • Forbes / CDC / Mayo Clinic (examples of authoritative sites)
  • BlackHatWorld (referenced in Lily Ray’s early experience with buying links)

Podcast/promotion source mentioned

  • Compact Keywords / compactkeywords.com (brief ad content; author not named in subtitles)

Original video