AI Overviews Hit 25%+ With External Links: What Peec’s Chart Means for B2B SaaS

Google can put more links in an AI Overview and still keep the buyer inside Search.

That is the tension in Peec AI‘s September chart, now circulating after Search Engine Roundtable covered it on September 23, 2026. The chart is useful. It is not a traffic forecast.

For B2B SaaS teams that score GEO success as “we got cited,” this week asks a sharper question: cited with what kind of outbound path, and did anyone leave Google?

---

What Peec’s chart claims

Malte Landwehr’s September 21 post stated that Google massively increased external links within AI Overview answers. The accompanying chart, as summarized by SER, shows the share of AI Overviews containing any external links moving from nearly zero to more than 25%.

What the public claim supports:

1. A large swing in link prevalence inside Overview answers (binary: answer has at least one external link).
2. Timing that sits after Google’s earlier 2026 work to make generative answers more link-rich (Further Exploration, inline citations, source previews).
3. A vendor-observed series, not a peer-reviewed causal paper.

What the public claim does not yet nail down in the open thread:

• Exact sample of queries, countries, devices, or crawl cadence.
• Mean links per Overview vs share of Overviews with one or more links.
• Whether those destinations are publisher URLs, Google Search refinements, or AI Mode follow-ups.
• Click-through rate on the new links.

Barry Schwartz at Search Engine Roundtable was careful on feel vs data: the chart “doesn’t feel right” against the broader zero-click mood, but he still published it. He also disclosed a small investment in Peec AI. Read that as transparency, and as a reason to replicate before you brief a CMO.

---

How this sits next to Overview → AI Mode leakage

The same news window includes a different Google test: normal-looking Overview controls that open AI Mode instead of publisher pages (SER Roundtable Sep 22; DerivateX write-up on citation ≠ click).

Those stories are not opposites. They stack:

Signal	What moved	What it does not prove
Peec chart	More Overviews show external-looking links	More organic sessions or demos
Overview → AI Mode test	Some “links” keep the next step on Google	That every chip is a dead end
Wang et al. field experiment	Forced AI Mode cut external CTR by 18.8 pp vs current Search	A universal site-level traffic forecast

If you only track “links in the Overview,” Peec’s chart looks like a publisher win. If you also track destinations and sessions, you may still be watching Google thicken the answer UI while rationing exits.

---

Why this matters for B2B SaaS pipeline

1. Link prevalence is a UI metric, not a pipeline KPI

A jump from ~0% to 25%+ of Overviews containing external links changes how the box looks. It does not automatically change demo starts. Pair it with sessions from Overview-prone URLs and with recommendation share vs citation share.

2. Search Console still will not finish the story

The GSC generative AI report gives impressions for AI Overviews and AI Mode. It still withholds AI-only clicks and the queries that triggered the answer. Rising Gen AI impressions plus Peec-style “more links” can coexist with flat referral.

3. Zero-click strategy still applies

DerivateX’s zero-click search strategy already treats citation and naming as pipeline inputs that often skip the session. More link chips raise the ceiling on exits. They do not remove the need for decision residue: branded search, CRM “how did you hear,” and shortlist naming on ChatGPT, Perplexity, Claude, and Gemini.

---

Operator playbook this week

Layer 1: Replicate Peec on your money prompts

Pick 20 commercial prompts your buyers actually ask. For each Overview:

1. Count whether any external destination appears (publisher URL vs Google-owned surface).
2. Click every control once. Log destination: site, Search refinement, or AI Mode.
3. Capture a screenshot set weekly so a UI change is visible without waiting on a vendor chart.

Layer 2: Wire three Google lines in analytics

1. GSC Gen AI impressions by URL (visibility).
2. Sessions / conversions on those same URLs (referral).
3. Branded search + CRM AI self-report (decision residue).

If Layer 1 rises because Overviews look denser with links, and Layer 2 stays flat, you won the box, not the visit. That can still be fine if Layer 3 moves.

Layer 3: Make the rare exit worth it

When a buyer does leave the Overview for your page, the first screen has to finish what the summary compressed:

• Pricing bands, packaging limits, SSO / security constraints, integration truth.
• A comparison table the model cannot fully flatten without losing buying criteria.
• A demo or trial path that matches the prompt class.

Thin “we got cited” landers waste the exit Peec’s chart implies is more common.

---

Worked example: mid-market RevOps SaaS

Priority prompts: “best RevOps platform for Series B,” “HubSpot RevOps alternatives with Salesforce sync,” “pipeline hygiene software for B2B.”

Weak read of Peec’s chart: declare Overview traffic “fixed” because 25%+ of answers now show links.

Better read:

1. Re-check those three prompts daily for a week. Log link count, destination type, and whether follow-ups stay in AI Mode.
2. Compare Gen AI impressions on money URLs to sessions and demo starts from the same URLs.
3. Invest first in extractable constraints and accurate third-party facts (G2, docs, integration directories), not another Overview-flavored fluff paragraph.

That turns a viral chart into a measurement queue.

---

How DerivateX reads “more external links”

1. Treat link prevalence as a UI change, not a traffic guarantee.
2. Always classify destinations: publisher, Google Search, or AI Mode.
3. Keep Gen AI impressions, outbound sessions, naming share, and CRM AI discovery as separate lines.
4. Assume vendor charts need replication when the publisher has a stake in AI-visibility tooling (Peec) or a disclosed investment (SER).

If you want a baseline that separates Overview link presence from naming and referral reality, start with a free AI visibility audit. Engagement options stay on pricing. For adjacent measurement, see the GSC generative AI report guide, zero-click pipeline strategy, and Overview links that open AI Mode.

---

FAQ

What did Peec AI report?

That Google sharply increased external links inside AI Overview answers, with the share of Overviews containing external links rising from near 0% to over 25%, per Malte Landwehr’s September 21 chart and SER’s September 23 write-up.

Is this independent research?

No. Peec AI sells AI-visibility products. Barry Schwartz disclosed a small investment in Peec when covering the chart. Replicate on your own query set before board reporting.

Does more links mean more traffic?

Not by itself. Link prevalence is not click-through, and some Overview controls still open AI Mode instead of websites.

How is this different from the Overview → AI Mode story?

Peec is about how often outbound-looking links appear. The AI Mode test is about where some of those controls actually go. Measure both.

What should we change this week?

Manual audit of 20 money prompts for link presence and destination, wire Gen AI impressions next to sessions, and harden first screens for the rare exit click.

Should we celebrate this as a publisher win?

Celebrate only after your replication shows real publisher destinations and after Layer 2 (sessions) or Layer 3 (CRM / branded) moves. UI density without exits is still a Google surface win.