I watched Search Engine Land’s “Is the Internet Broken? How Google and LLMs Changed the Web Forever” and you should too because it names the real shift: AI answers now sit between your content and your customer, and a lot of informational clicks aren’t coming back. 🧭
Here’s why that matters and what I’d actually do about it right now.
If you run content or SEO, you don’t fix this with more explainers. You reset targets, instrumentation, and formats so assistants cite you or hand off to you. This is the playbook I’m putting in motion.
// My reaction
Frustration: explainers won't save you
- AI answers are reshaping discoveryTreat informational traffic as at-risk; shift your roadmap toward citeable assets, transactional tasks, and defensible intent. Instrument it, don’t guess.
Why this argument matters for working SEOs
- The web is saturated with LLM-spun sameness. Google and assistants answer inline, and a growing share of queries end in clickless SERPs. 📉 Studies disagree on exact rates, but zero‑click behavior is structurally high (e.g., SparkToro/Datos 2024) and AI surfaces contribute to fewer open‑web clicks for many informational queries (https://sparktoro.com/blog/2024-zero-click-search-study-for-every-1000-us-google-searches-only-374-clicks-go-to-the-open-web-in-the-eu-its-360/). (sparktoro.com)
- Triggering is uneven and moving. Google’s AI Overviews (AO) hit broad U.S. rollout on May 14, 2024, then tightened after quality issues; measured prevalence varies widely by dataset and time window (9to5Google; BrightEdge) (https://9to5google.com/2024/05/14/google-search-ai-overview-rollout/, https://help.brightedge.com/resources/weekly-ai-search-insights/ai-overviews-one-year-presence-size-citing). See also Google’s own post acknowledging adjustments after early errors (https://blog.google/products-and-platforms/products/search/ai-overviews-update-may-2024/). (9to5google.com)
- The training loop risk is real: models trained on model‑generated content degrade over time (“model collapse”), which pushes you toward original, source‑quality material if you want citations (Nature 2024) (https://www.nature.com/articles/s41586-024-07566-y). (nature.com)
- For local and branded demand, the shift is different: assistants still route to businesses to complete tasks (bookings, calls). That’s winnable via structured actions and clean entities (Reserve with Google partners, Google docs) (https://www.google.com/maps/reserve/partners, https://support.google.com/reserve/answer/9172607?hl=en, https://support.google.com/localservices/answer/9905231?hl=en). (google.com)
| Action | Primary metric |
|---|---|
| Identify AI-risk terms | Queries with AI answer presence |
| Citeable content sprints | Citations earned / Links earned |
| Task-focused pages | Actions completed (leads, bookings) |
| Local entity cleanup | Calls, direction requests, GBP views |
What to prune
- generic how-tos with no data
- copycat listicles
- orphaned thought pieces
What to ship
- transactional tools and checklists 🛠️
- original tests, pricing, and methodologies
- citeable definitions, tables, and templates
Is Google showing AI Overviews on everything now?
No. AO presence varies by topic/intent and has fluctuated since launch; credible snapshots disagree widely because of different datasets and windows (e.g., BrightEdge). Validate against your own queries.
Will structured data make AO cite me?
No. There’s no AO-specific schema and markup isn’t required for generative features. Schema helps understanding/eligibility, but citations hinge on being a source with specific, extractable information (Google docs).
Should I block AI in robots.txt to stop AO from using my content?
Be careful. You can control some AI usage (e.g., Google-Extended for Gemini training; Search Console toggle for generative features), but blocking Googlebot to avoid AO also kills normal indexing/ranking. Test before you torch visibility.
Is zero-click proof that SEO is dead?
No. Zero-click is high and rising, but search demand is still massive. The work shifts to assisted conversions, citations, and task handoffs — not just sessions.
Do AO links ever send traffic?
Sometimes. Google says AO includes supporting links, and independent analyses show reduced but non-zero CTR that varies by query. Plan for lower click density and higher brand/assistant exposure.
What I’d do in the next 90 days
1) Measure your AI exposure and protect what pays
- 🔎 Segment pages by intent and revenue contribution. Identify which URLs actually drive leads, sales, or sales‑assisted sessions. Guard those first.
- Track AI Surface Risk by query: fully answerable vs. must-click. Use your AI Visibility Checker (tools/ai-visibility) weekly to spot terms drifting into AI Overviews or other assistant answers. Cross‑check against Search Console’s generative features reporting where available (https://support.google.com/webmasters/answer/16908024?hl=en, https://support.google.com/webmasters/answer/16984139?hl=en, https://developers.google.com/search/blog/2026/06/gen-ai-performance-reports). (support.google.com)
- Adjust KPIs: de‑prioritize raw TOFU sessions; focus on assisted conversions, AO/assistant citations, and action completions. If you need a measurement pattern, I outlined a defensible proxy in Track AI Search Traffic in GA4 — How to Build a Defensible Proxy (/blog/track-ai-search-ga4-proxy).
- Pro tip: Validate AO presence against your own SERP fetches. BrightEdge and others show conflicting AO coverage ranges because methods differ; your vertical view matters more (https://help.brightedge.com/blog/10-observations-AI-Overviews-June2024). (help.brightedge.com)
2) Make content citeable, extractable, and worth linking to
- 🧪 Publish first‑party specifics: pricing actuals, test designs/results, screenshots, checklists, legal/regulatory frameworks, before/after evidence. Models quote specifics, not fluff. I break down patterns in Build Content for AI Search: Answer‑First, Citeable, and Extractable (/blog/build-content-ai-search).
- Structure for machines and people: tight definitions, step‑by‑steps, and small tables near the top. Add schema where it clarifies entities, actions, and outputs but don’t expect schema alone to drive AO citations. Google doesn’t offer AO‑specific markup and says structured data isn’t required (Google docs) (https://developers.google.com/search/docs/fundamentals/ai-optimization-guide). (developers.google.com)
- Target AO citation patterns: AO often cites beyond the “usual top 10,” which is your opening if you bring net new information (BrightEdge observations) (https://help.brightedge.com/blog/10-observations-AI-Overviews-June2024). (help.brightedge.com)
- Ship supporting artifacts that travel well in summaries: one‑sentence definitions, numbered procedures, and small comparison tables. Reference them across pages so they earn consistent mentions and links.
3) Win tasks, not trivia: aim at “do” intents
- 🛠️ Build pages that trigger handoffs: calculators, instant quotes, inventory finders, schedulers, refund or eligibility checkers. If the assistant can summarize but not execute, give it a clear action to offer.
- 📲 Expose actions cleanly with labels and schema, then test them in assistants. Mirror those with GBP appointments and any Reserve with Google or Local Services integrations you qualify for.
- Make your UX obvious at a glance. Prominent CTAs, short forms, and autofill where possible. If a bot or a busy human scans your page for five seconds they should know the next step.
- Keep a tight entity footprint. Canonical names, addresses, categories, and product identifiers aligned across your site, GBP, and major directories so assistants can trust your match and send the click.
Takeaway
The internet isn’t broken, it’s reweighted. Informational commodity content is losing surface area. Task completion and source‑quality material are gaining it. Reset your roadmap, instrument risk, and publish work that assistants cite or hand off to. That’s how you protect revenue and still grow in AI search, which is exactly where SEO, GEO, and AI now meet. 🤝
// My reaction
