B2B Search & Visibility Digest: Week of August 17, 2026

The Big Picture Google Ads forced the week's biggest operational deadline: a mandatory Demand Gen/Performance Max bidding change landed August 17, and Dynamic Search Ads began their phase-out in favour of AI Max. Meanwhile the organic side kept fragmenting: ranking volatility spiked without a confirmed core update, Google pushed back on claims that Reddit gets preferential treatment, and fresh data shows AI Overviews are now pulling far fewer citations straight from the classic top-10 SERP than a year ago. For B2B teams, the throughline is the same one that's defined 2026 so far: paid and organic visibility increasingly depend on machine-readable signals (structured data, llms.txt, CRM-fed conversion data) rather than page-one rank alone.

Google AI Mode & AI Overviews Fewer citations are coming from the classic SERP Ahrefs' latest analysis of 863K keyword SERPs and roughly 4 million AI Overview URLs found that Google is pulling a shrinking share of AI Overview citations directly from the organic top 10, down to roughly 38% of citations sourced that way, versus about 76% a year earlier. That means ranking well organically is no longer a reliable proxy for showing up in an AI Overview; Google is increasingly drawing on a broader, separately crawled pool of sources.

Key stat: Only ~38% of AI Overview citations now come from a page's original top-10 ranking position, down from ~76% in mid-2025 (Ahrefs). Google disputes Reddit favouritism, cites AI reshuffle Amid a week of unusually volatile rankings, Google publicly stated that Reddit "gets no special preference" in search rankings or AI features, after site owners noticed Reddit threads losing visibility. The clarification landed alongside reports of a broader Alphabet AI reorg, with some teams moving out of DeepMind into core Google search and ads product groups, a signal that AI Mode and AI Overviews development is being pulled closer to the core Search org.

No confirmed core update, but real turbulence Google's Search Status Dashboard showed no ranking, indexing, or serving incident in the first half of August, yet third-party rank trackers spiked starting August 1 and stayed elevated through the following week. Some SEO vendors are attributing the volatility to an unconfirmed update targeting "weak originality and thin evidence" pages, worth watching if your B2B content relies on aggregated or lightly sourced material.

Sources: AI Overview Citations Report, Google says Reddit gets no special preference in search, Reddit loses search traffic as Google denies ranking preference, Search Engine Roundtable: Wild Ranking Volatility, Reddit, Executive Shakeup

GEO / AEO / AIO AI referral traffic is still small but converts far better New publisher-side data shows AI chat referrals make up only about 1.1% of visits following an AI chat session, most AI-influenced traffic still arrives via direct or regular search rather than an inline link. But conversion data continues to favour AI-referred visitors: ChatGPT referral traffic is converting around 15.9%, Perplexity around 10.5%, and Claude around 5%, all well above typical organic-search conversion rates. For B2B pipelines specifically, Perplexity's smaller volume (15 to 20% of AI referral traffic) is reportedly delivering the highest ROI per citation of any AI platform.

Key stat: ChatGPT's share of B2B AI referral traffic fell from roughly 89% eight months ago to about 63% this quarter, as Gemini and Perplexity pick up share. ChatGPT's dominance is eroding Gemini grew 231% year-over-year and has overtaken Perplexity as the number two AI traffic source; Perplexity's own US AI-traffic share slipped from ~11.4% to ~6.9% over the same period. For B2B teams building a GEO program, this argues against optimizing for a single engine. Gemini's growing footprint (it's now the default model behind AI Overviews globally) makes it a citation target in its own right.

Track citations across ChatGPT, Gemini, Perplexity, and Claude separately: engine share is shifting monthly, not annually. Weight GEO effort toward E-E-A-T signals and third-party citations/digital PR, which multiple studies continue to tie to citation odds.

Sources: AI SEO News Recap - August 2026, ChatGPT's AI Referral Share Fell From 89% to 63%, ChatGPT vs Perplexity for AI Visibility in 2026, Search Engine Land: SEO news library

Traditional SEO & Organic Search B2B buyers still start more searches on Google than on AI Despite the GEO narrative, Google still owns roughly 85% of global search market share and search overall drives about 76% of B2B website traffic, a reminder that classic organic SEO isn't being displaced so much as supplemented. Original research continues to outperform generic content: B2B SaaS sites publishing proprietary research grew top 10 keyword rankings by ~21% and organic traffic by ~19% on average, ahead of sites relying on repurposed or AI-generated content.

Key stat: 51% of B2B software buyers now say they start vendor research with an AI chatbot more often than with Google, but Google still delivers the largest single share of overall search traffic to B2B sites. Search Console keeps expanding beyond Google Search itself Google added platform-level reporting in Search Console covering X, Instagram, YouTube, and TikTok, letting site owners see which queries drive impressions and clicks to their social and video profiles, useful for B2B teams treating social search as part of the discoverability mix, not a separate channel.

Sources: August 2026 SEO & GEO News Recap - ROI Revolution, Search Engine Land: PPC & SEO news library, B2B SEO in 2026: What's Changed and What Still Wins

New Protocols & Standards llms.txt inches toward formal standardization The llms.txt spec, the proposed file that tells AI crawlers how to interpret a site, was updated again on August 10 by original author Jeremy Howard, and a June 2026 W3C draft has already proposed formalizing the spec: locking the file to the domain root (no subpath variants), adding a version header for crawler compatibility checks, and restricting the body to a strict markdown subset. Adoption data is starting to back the effort: Princeton's GEO-bench research found sites with a well-structured llms.txt earned 23% more AI citations across ChatGPT Search, Perplexity, and Google AI Mode versus matched sites without one.

Key stat: Sites with a well-structured llms.txt file earned 23% more AI citations than matched control sites (Princeton GEO-bench). FAQ schema officially gone Google's FAQ rich results, retired from Search on May 7, lost their Search Console reporting filter in June, and API support for the deprecated FAQ report ends this month (August 2026). If your team still has FAQPage markup deployed for SEO value, it's now purely a GEO/citation play, not a SERP-feature play; Google has also reiterated that no new special schema is required to appear in AI Overviews or AI Mode.

Sources: llms-txt.org spec, W3C llms.txt standard proposal - Pravin Kumar, Structured Data After I/O 2026: Schema Cheat Sheet, Schema.org Structured Data and AI Search - MOCOBIN

Google Ads Updates Mandatory Demand Gen/PMax bidding change landed August 17 Google updated bidding within Demand Gen line items (and by extension Performance Max) on August 17 to make target-based strategies (Target CPA, Target ROAS, Target CPC) track more strictly to the bid target, including when budgets change. Campaigns that had been quietly overperforming their targets should expect performance (and often cost-per-conversion) to shift back toward the number actually set. This is not opt-in; B2B teams running PMax or Demand Gen for lead gen should re-check target settings before assuming a performance drop is a targeting or creative problem.

Key stat: The August 17, 2026 bidding update applies automatically to all Demand Gen line items using target-based bid strategies, no opt-out. Dynamic Search Ads is being retired in favor of AI Max Google confirmed DSA campaigns will be phased out and upgraded to AI Max for Search. Voluntary migration tools rolled out this week to port historical settings into new standard ad groups; campaigns using Automatically Created Assets or campaign-level broad match will be auto-upgraded starting September 2026, with the rest following in a later wave. For B2B lead-gen accounts, AI Max's main relevance is enhanced conversions for leads (passing hashed CRM data back to Google) and value-based bidding across MQL/SQL/closed-won stages; B2B advertisers using these features together are reportedly seeing materially better ROI than AI Max used on its own.

Turn off URL expansion in AI Max if you rely on dedicated landing pages, otherwise traffic can land on non-converting pages. Feed CRM-verified MQL/SQL/closed-won data back into value-based bidding rather than optimizing on raw lead volume.

Sources: Google: We're upgrading Dynamic Search Ads to AI Max, Demand Gen bidding changes - Google Help, Performance Max bidding change, August 17 - pmax.online, Every Google Ads Change That Matters in 2026

Suggested Actions This Week Audit Demand Gen and Performance Max campaigns against the August 17 bidding change before assuming any CPA/CPL shift is a creative or targeting issue. If still on Dynamic Search Ads, start the voluntary migration to AI Max now: the auto-upgrade for ACA/broad-match campaigns begins in September. Publish or refresh an llms.txt file and confirm it reflects current site structure; early data ties it to a meaningful bump in AI citation rates. Don't over-index on top 10 organic rank as a proxy for AI Overview visibility: with only ~38% of citations now sourced from the classic SERP, treat GEO as a distinct workstream. Diversify GEO tracking beyond ChatGPT: Gemini's rapid growth and Perplexity's high B2B conversion rate both warrant dedicated monitoring. Remove any remaining FAQ Page schema built purely for the old FAQ rich-result feature; redirect that effort toward Organization, Product, and Course schema, which show stronger ties to AI citation.