- Bold ledeAI “reports” are expanding. Measure displacement and citation lift before you change a single page.
- Bold ledeAI‑generated store rating reviews are live in more surfaces. Clean, consistent review inputs prevent messy summaries.
- Bold lede“Crusty old links” still cast shadows. Audit by era/type; prune what you control; disavow only with a documented pattern.
- Bold ledeThis week’s sprint: baseline AI exposure, harden review ops, and run a legacy link audit. Inputs first, always.
- 01Build a baseline
Implement the GA4 proxy and log queries where AI summaries appear.
- 02Identify evidence gaps
Map claims to sources; add missing citations and schema for entities, products, and FAQs.
- 03Test snippet intent
Rework sections to answer succinctly with scannable facts and links to depth.
- 04Monitor citations weekly
Track changes in AI mentions vs. classic SERP clicks to guide content sprints.
AI Overviews compress. Your job is to reduce ambiguity. Tighten evidence, markup, reviews, and links so machines have no choice but to pick you.
Can I see AI click and query data in the new GSC AI report?
Not currently. The Generative AI performance report exposes impressions by page, country, device, and date. No clicks or queries are documented. Use it for presence and pair it with on‑site analytics for impact (help).
Can I “opt out” of AI training with the new Search Console control?
The control governs inclusion/links in generative AI features (AI Overviews, AI Mode, Discover). It doesn’t control model training; Google points to Google‑Extended for training limits (control scope).
Do AI Overviews always kill clicks?
Mixed. Google says AI Overviews link out and can lead to “higher‑quality” clicks; independent studies show displacement varies by query class. Measure your classes, don’t generalize (Google explainer).
Are AI‑written Google reviews allowed if the customer approves them?
No. Reviews must reflect genuine, first‑hand experiences; fake or misleading content is prohibited. In the U.S., the FTC’s rule bans fake reviews and enables penalties (policy; FTC rule).
Should I mass‑disavow legacy spam?
Only with a documented pattern and risk. Misuse can hurt performance. Start by pruning influenceable links; if a clear pattern persists, then disavow carefully (Disavow Help).
Barry Schwartz just posted a tight roundup worth your time: It’s New 9/1. Why I care is simple it hits the three levers actively shifting search exposure right now AI summaries, local and review surfaces, and link risk 🧭
I’m not here to echo headlines. I’m here to translate them into moves you can actually ship this week without torching your roadmap.
Here’s the point by point on what matters and why.
- AI “reports” are expanding beyond early tests. Treat them like displacement you must measure, not a channel you own. 🌍
- AI-generated store rating reviews are showing up. Your review inputs will be summarized by machines curb volatility now. 🛍️
- “Crusty old links” are back in the discourse. Legacy link risk still exists; don’t let ancient junk define your profile. 🧱
- This week measure AI impact, harden review ops, and audit legacy links. Prioritize inputs you control. 🔧
What Barry surfaced and why it matters
- AI “reports” Google’s AI generated overviews and summaries surfacing more broadly means less blue link real estate and more answer first UI. That’s displacement risk and a citation opportunity at the same time, measure both or you’ll chase shadows.
- AI generated store rating reviews signal Google will keep synthesizing user feedback. If your inputs review mix, attributes, owner responses are messy the summary will be messy and you’ll hate how it reads.
- The chatter on old links is a reminder link equity is path dependent. Bad neighborhoods and lopsided anchors from years ago can still skew how machines model your site and if you ignore it that’s on you.
Bottom line: focus on inputs that train these systems your content evidence, structured data, and reputation signals.
How to measure the AI impact before you react
If you don’t quantify AI displacement vs citation lift you’ll chase ghosts and burn cycles.
- Start with a defensible proxy for AI sourced traffic. I outlined one here: Track AI Search Traffic in GA4 — How to Build a Defensible Proxy. 📊
- Check where you’re cited today. Use the AI Visibility Checker to spot which queries and pages show up in AI summaries.
- Strengthen evidence and markup on pages likely to be summarized. My rule one primary claim per section, backed by a source or first party data, and explicit entities via schema. The Schema Opportunity Analyzer will surface low hanging markup gaps.
- If you need a primer on earning citations, read: AI Search vs SEO: How to Get Cited by AI Overviews.
- 01Build a baseline
Implement the GA4 proxy and log queries where AI summaries appear.
- 02Identify evidence gaps
Map claims to sources; add missing citations and schema for entities, products, and FAQs.
- 03Test snippet intent
Rework sections to answer succinctly with scannable facts and links to depth.
- 04Monitor citations weekly
Track changes in AI mentions vs classic SERP clicks to guide content sprints.
Local and reviews tighten the inputs
When Google synthesizes reviews, it compresses your reputation into a paragraph. You want that paragraph to read like your best customer not your loudest outlier.
// My reaction
Reviews distilled — handle with care
- Balance the review mix recency, rating, and topics. Encourage specifics that align to your key attributes service lines, timelines, locations. Never incentivize just ask better and be consistent.
- Respond like a human, not a template. Owner replies are inputs too. Machines summarize tone and they will quote your canned lines back at you.
- Normalize your product and service attributes across your site, GBP, and key directories. Consistency prevents odd AI mashups that make you look sloppy.
- Add first party proof to service pages before after, process, guarantees so AI summaries have facts to pull, not fluff.
Links audit, don’t time travel
I’m not nostalgic about links. If Barry’s roundup reminded you of “crusty old links,” take the hint legacy baggage can still distort.
- Pull a fresh link export. Group by era and source type. Look for old networks, sitewide footers, or directories that aged poorly and note anything you’d be embarrassed to show a client.
- Prune what you can influence your own properties, partner pages. For the rest, document patterns and watch anchor ratios so you see the risk before someone else does.
- Disavow only with a clear pattern and a paper trail. Most sites need restraint, not theatrics, I’ve cleaned up more messes caused by performative disavows than by bad links.
Goal: a link graph that looks earned, current, and coherent with your topical focus not a museum of 2014 tactics.
Takeaway
Barry’s video isn’t drama; it’s a nudge. AI summaries are expanding, reviews are being machine written, and old links still cast shadows. Measure displacement, shore up reputation inputs, and clean the legacy noise. Then iterate and keep receipts.
If you run SEO, GEO, or AI driven content ops, this is the job reduce ambiguity in your signals so machines have no choice but to pick you.
