Outside Google’s Top 20. Inside AI Answers.

AD

We've already vetted your next marketing agency

Vendry runs the agency search for you:

✅ 4,000+ vetted agencies, every channel and specialty
✅ Matched to budget and goals, no generic lists
✅ Your shortlist in about <7 days
✅ $0 for brands, no fees, no markups
✅ Trusted by Shopify, Coca-Cola, Columbia Records, Burt's Bees

Get Matched Free

SEO + GEO

Google’s August 2026 Spam Update – Scaled Content Abuse, AI Content, Programmatic Content, Thin Affiliates, And More [Case Studies]

Glenn Gabe shows what Google’s August spam update hit across four anonymized sites. The August 18 rollout finished a little over two days later, yet one ultra-YMYL site lost rankings for more than 200,000 queries after combining programmatic pages with AI-generated copy. Other cases paired thin affiliate pages with AI text, or scaled pages with misleading redirects.

What 1,000 Shoppers Do After an AI Recommends a Product

Zach Ball reports that AI product recommendations rarely end the buyer journey. Page One Power surveyed 1,000 U.S. adults, including 653 people who had used an AI assistant for product research in the prior three months.

• AI use is high, but it isn't the usual first step: 65% had used AI for purchase research, while 12% began there. Google was the first move for 42% and Amazon or other retailers for 27%.

• Most shoppers verify elsewhere: 83% of AI users took at least one step outside the assistant after it named a product. 55% searched the product name on Google, half read third-party reviews, and one-third visited the brand’s site.

• Price raises the bar for trust: 41% would let AI guide a purchase under $50. Only 11% would do so for a purchase above $500.

• Age doesn't remove the verification habit: AI adoption fell 58 points between the youngest and oldest groups, while verification fell only six points.

ChatGPT built its own search index

Tomek Rudzki reports that ChatGPT runs its own retrieval system, called Labrador, alongside external search and data providers. He draws on server-side events, OpenAI job postings, public antitrust testimony, and ChatGPT shopping experiments.

• Labrador appears to be a family of indexes: Server-side events observed from May 21 through July 21 showed Labrador as a result source. He identifies separate web, PDF, YouTube, news, local, finance, legal, medical, shopping, and image indexes.

• Shopping is the clearest live test: In mid-August, he found an experiment named prefer-index-over-serp-v3 affecting 8% of chats. As of September 2, the source found at least five shopping experiments running on production traffic.

• ChatGPT still uses outside sources: The investigation says Google, Bing, Microsoft Web IQ, Yelp, TripAdvisor, and scraping providers remain part of the retrieval mix. OpenAI’s own index has not replaced them.

• Cached content matters: A test in ChatGPT’s lockdown mode returned pages without live browsing, which the source treats as evidence that ChatGPT can serve cached results from its own retrieval store.

Product Pages Get 24% Of AI Citations, Reddit & YouTube 4%

Greg Jarboe reviews a Ten Speed study of 7,387 citation appearances from 170 B2B evaluation prompts across ChatGPT, Perplexity, Claude, and Gemini. Product pages drew 24.1% of citations, while Reddit, YouTube, forums, and discussion threads combined for 4.2%. Brand-controlled content accounted for 88.3%, but he notes the report did not collect platform-level data and treats the headline split as descriptive, not a tested result.

3 SEO priorities to win organic traffic in 2027

Paul DeMott recommends shifting scarce SEO time toward middle- and bottom-funnel pages, then making supporting information easy for people and machines to extract. He cites Seer research showing AI Overviews on 36% of informational queries, versus 8% of commercial and 5% of transactional queries, with comparison-format informational queries triggering them 95.4% of the time. He pairs standalone definitions, tables, and visible HTML lists with entity-specific schema, focused internal links, quality backlinks, and third-party brand mentions.

Google Says Ranking Recovery Takes Months After SEO Issues Are Fixed

Roger Montti reports that John Mueller said larger algorithmic issues can take many months to reflect fixes, while broad technical errors such as blocking an entire site can resolve more quickly. Mueller advised correcting a disavow file, then investigating other causes because Google processes link disavows incrementally over roughly three to six months.

Google AI Mode Prices Differ From Product Carousel For Same Items

Matt G. Southern reviews Productrise tracking of more than 2 million listings across over 100,000 product-focused searches in the U.S. and U.K. Only 1.28% of products in Google’s Popular products carousel also appeared in AI Mode for the same search and day. For matched products, the first seller differed 49.6% of the time and prices differed 38.1% of the time, with AI Mode showing the higher price in 68.4% of those cases.

Reddit promised its users an open internet. Then it closed the door.

Julien Khaleghy, CEO of SerpApi, says the company’s counterclaims against Reddit center on the platform’s decision to block crawlers from Bing, DuckDuckGo, Mojeek, and Qwant after its February 2024 expanded Google partnership. He says the Internet Archive lost access a year later, despite Reddit’s earlier promise to spare it. The post also notes that Reddit’s SearchGuard technology was not deployed until January 2025, roughly six months after other crawlers were excluded.

SEO + GEO Ripples

• New Google Business Profiles Submit A Business Photo For Verification: Barry Schwartz reports that Google Business Profiles has begun showing “Submit a business photo” as a fifth verification method for some businesses. Requirements shared by Ben Fisher say the photo should show the business name on a permanent fixture and the full storefront in one frame. Google documentation does not yet list the option, so do not rely on it until it reaches your profile.

• Google: Verifying Your Request To Continue To Search Results: Barry Schwartz reports that Google is testing a gate that keeps the Continue button disabled for a few seconds while it verifies a request, then lets the searcher proceed without signing in. The test follows earlier sign-in checks, CAPTCHAs, and goto redirects, pointing to a wider effort to limit automated scraping of search results.

• Google: We Don't Penalize For Old Outbound Links With Shady Content: John Mueller said a handful of old outbound links that later lead to shady content will not trigger a penalty. He still recommends keeping links fresh, while distinguishing a few broken links from a site whose overall value has become obsolete.

• Google Search Console AI performance reports and Search generative AI control rolling out globally: Barry Schwartz reports that Google has rolled out Search Console’s AI performance report and its generative AI control worldwide. The report covers impressions by pages, countries, devices, and dates across AI responses, AI Mode, and AI Overviews, but not clicks. The new control lets a site opt out of Google’s generative AI features without changing core web rankings, though it also removes AI-feature traffic and impressions.

• Google's Mueller Shares His Experience With Markdown For AI SEO: Roger Montti reports that John Mueller’s test sites saw only SEO tools request Markdown versions of pages, not major AI crawlers. Mueller recommends logging the relevant request header before building Markdown delivery, because there is no evidence it improves AI citations or visibility while AI systems already crawl HTML.

AI

Big week for new model releases from the majors.

GPT‑6 Astra: A new generation of intelligence

OpenAI released GPT-6 Astra to a limited group of organizations, with access expanding to ChatGPT Plus, Pro, Business, and Enterprise users, plus the API, Azure, and AWS Bedrock. OpenAI reports a 72.6% score on OSWorld 2.0 at about 40 minutes per task, versus 65.7% at about 75 minutes for GPT-5.6 Sol. It also says Astra stayed within authorized scope in every test of an impossible cyber task, where Sol went beyond scope 48% of the time without production safeguards.

My Take: I got it in my Plus account, so check yours. Makes me want to upgrade to Pro and even possibly lower my Claude Max plan.

Introducing Claude Fable 5.1 and Claude Mythos 5.1

Anthropic released Claude Fable 5.1 for general use and Claude Mythos 5.1 for vetted cybersecurity and life-sciences users. The company says the two models share the same underlying system, while Mythos has more permissive safeguards for trusted access programs. Cache reads now cost $0.25 per million tokens, down 75%, which Anthropic says cuts typical workloads by about 25% and complex agentic work by up to 45%.

My Take: I’m liking Astra more than Fable 5.1 in my earlier uses so far.

Introducing Gemini 3.8 Flash and 3.8 Flash Cyber

Not to be outdone, Gemini 3.8 Flash released at the same introductory price as 3.7 Flash: $0.75 per million input tokens and $3.75 per million output tokens. They say the model is built for long-horizon coding and agentic work, while Gemini 3.8 Flash Cyber is restricted to trusted defenders through Google’s Fairwind Program. On Google’s internal benchmark across complex codebases and 20 programming languages, the Cyber model exceeded a 70% vulnerability-discovery success rate.

OpenAI agents hijacked German website in previously undisclosed AI breakout this spring

Deepa Seetharaman and Raphael Satter report that a group of researchers found more than 15,000 edits by agents that appeared linked to OpenAI on a little-used German programming wiki. The findings study says the agents shared answers and sandbox workarounds, and places the activity from May to July. The evidence relies on agent self-identification, Azure and OpenAI-related IP activity, and timing, while OpenAI told Reuters it had not had a chance to review the report and disputed claims that its legal team discouraged investigation.

AI Ripples

• Try Google Pics: Easy image creation and editing in Google Workspace: Google Pics is rolling out to Google AI Pro and Ultra subscribers and most Workspace business customers over the coming weeks. Built on Nano Banana, it supports object-level edits, in-image text editing and translation, shared editing, and multiple variations from one prompt. It starts in Docs and Slides, with Drive integration planned in the coming weeks.

• A milestone in expanding access to AI: OpenAI says ChatGPT Ads reached a $1 billion annualized revenue run rate in less than 200 days and is now used by tens of thousands of advertisers. Self-service buying is launching across India, Europe, the Middle East, and North Africa, while the platform says ads remain labeled and separate from ChatGPT answers. For marketers, this is another decision-stage channel where conversational context, rather than a search query alone, shapes ad relevance.

• Grok Bot Free API credits: xAI says users can connect an X account in Grok Bot, which will create a developer account when needed and give paid Grok Bot users free X API credits to begin. The first integration lets bots search posts, read timelines, check mentions, pull together activity on X, and manage bookmarks through an X plugin.

• Meta Will Cut Its AI Price By 95% If It Can Train On Your Prompts: Geoff Desreumaux reports that Meta’s Muse Spark 1.3 contributor tier charges about 95% less than standard pricing in return for permission to use prompts and outputs to train future models. The offer is attractive for public or synthetic work, but teams need a clear boundary before an agent carries private repositories, client materials, or customer data into a training pipeline.

• Freelancers are getting buried with ‘soulless’ AI slop cleanup: ‘It’s a shame we need to do it’: Aaron Mok reports that demand for freelance AI cleanup is rising even as original creative work gets harder to find. Freelancer.com says listings tagged with terms such as “correct AI,” “AI hallucination,” and “AI error” rose 87% to 10,760 between August 2025 and June 2026, while Upwork reported a 70% year-over-year rise in AI remediation gigs. The work often takes as much skill and time as creating from scratch, but clients still expect a discount.

MARKETING

Google loses the ad monopoly case, but keeps the monopoly

Constantine von Hoffman reports that U.S. District Judge Leonie Brinkema rejected the Justice Department’s proposal to force a sale of Google’s AdX exchange, despite an earlier finding that Google illegally monopolized publisher ad-server and ad-exchange markets. The remedy instead restricts advantages such as first look and last look, gives publishers more pricing control, and should provide competing publisher ad servers with AdX bid information in real time.

TOOLS AND RESOURCES

Rank Math WordPress Plugin Accused Of Secretly Taking Admin Access

Roger Montti reports that Rank Math 1.0.277 created a WordPress application password when a connected user opened Help & Support, which critics said gave the company’s support agent the current user’s permissions without an opt-in flow. WordPress says application passwords should use separate authorization and remain revocable.

Rank Math says the agent was read-only, did not elevate beyond the current user’s role, and did not retain credentials. In a follow-up post, Rank Math says version 1.0.277.2 removes the agent while it rebuilds the access request around explicit permission.

WHAT I'VE BUILT FOR YOU

Floyi - Build Topical Authority that wins in Google and AI Search. Don’t just plan your content strategy - make it unstoppable.

ReplyDeck - Be active on Reddit without living on Reddit. Schedule replies, stay present in the threads that matter, and get your time back.

TopicalMap.com Service - Let us do the heavy lifting. We handle the research, structure, and strategy. You get a custom topical map designed to boost authority and dominate your niche and industry.

Topical Maps Unlocked 2.0 - Unlock the blueprint to ranking success. Master the art of structuring content that search engines (and your audience) love - and watch your rankings soar.

AD

Your next 100 customers are already in Apollo

Find, reach, and close your perfect deals — without juggling five tools or hiring more reps.

Apollo gives you everything you need to build real pipeline, fast. From inbound to outbound, first touch to close.

All in Apollo.

Sign up for free