Contents
- 1 🏛️ Official Updates
- 1.1 Evolving role of the index: From ranking pages to supporting answers
- 1.2 5 new ways to explore the web with generative AI in Search
- 1.3 5 gardening tips you can try right in Search
- 1.4 How frontier firms are pulling ahead
- 1.5 GPT-5.5 Instant: smarter, clearer, and more personalized
- 1.6 Advancing voice intelligence with new models in the API
- 1.7 Uber uses OpenAI to help people earn smarter and book faster
- 1.8 Testing ads in ChatGPT
- 1.9 Parloa builds service agents customers want to talk to
- 1.10 New ways to buy ChatGPT ads
- 2 🤖 GEO·SEO Highlights
- 2.1 Bing Reveals What Grounding Means For AI Search Visibility
- 2.2 How Does AI Get Its Information? Training Data, RAG, MCPs, and APIs Explained
- 2.3 What is Index Bloat? — Whiteboard Friday
- 2.4 Attribution gap in agentic search: how to close it
- 2.5 How to optimize for AI search results in 2026
- 2.6 On-Page AEO: 4 Writing Frameworks for Better AI Visibility
- 2.7 Agentic AI vs. Generative AI: What’s the Difference and Why It Matters
- 2.8 6 Generative Engine Optimization Benefits Every Marketer Should Know
- 2.9 SEO Migration Checklist: How to Switch Tech Stacks Without Losing Rankings (Developer Guide)
- 2.10 Keyword Clustering: How to Build a Topic Authority Strategy for 2026
- 2.11 Product SEO: 8 Strategies That Drive Demand for B2B & SaaS
- 2.12 The Agent Runtime Wars Have Begun. Is Your Website Ready?
- 2.13 Claude Skills for SEO and Marketing: What They Are and How to Use Them
- 2.14 The Reddit Earnings Story Most Marketers Missed
- 2.15 AI Companies Are Selling Heartwarming Ads – They’re Racing To Automate Your Job
- 2.16 Google’s Mueller Flags SEO Gaps In AI Vibe Coding
- 2.17 Google’s Quality Threshold Is Quietly Killing Scaled AI Content At Ranking
- 2.18 Google Answers Whether Preferred Sources Overrides Low-Quality Signals
- 2.19 How AI Will Transform PR’s Role in SEO Strategy Over the Next Two Years
- 2.20 How to Track ChatGPT Traffic for You and for Competitors
🏛️ Official Updates
Evolving role of the index: From ranking pages to supporting answers
We see the evolving role of the index: it now powers AI answers, not just ranked pages.
Traditional search optimized for documents. AI grounding optimizes for verifiable facts. We must measure factual fidelity, source quality, and freshness differently. Stale facts now produce wrong answers, not just lower rankings. Abstention becomes a valid outcome when evidence fails. This shift demands new index metrics for trustworthy AI responses.
5 new ways to explore the web with generative AI in Search
Google just dropped 5 new ways to explore the web with generative AI in Search.
I recommend you take note. These updates push users straight to trusted content. You now see article suggestions at the end of AI responses. News subscription links get a clear “Subscribed” label. That label boosted click-through rates significantly in early tests. You also get previews of community advice from forums and social media. The system shows the creator’s name or community name. Links appear right where you need them in the response. For publishers, Google offers a form to link subscriptions. This shift directly rewards original, authoritative sources. Update your SEO strategy to capture these surfaced links.
5 gardening tips you can try right in Search
As a Google Search optimization strategist, I recommend you test these 5 gardening tips immediately. Google’s AI Mode, Canvas, Shopping, and Search Live give you direct, measurable advantages for content visibility.
First, this official Google article proves AI-powered search now answers complex, visual, and local queries. AI Mode visualizes garden layouts from a photo. Canvas builds year-long planting schedules with one prompt. Shopping filters show “in stock nearby” inventory. Search Live identifies plant diseases in real time via camera.
We see concrete trends: “chaos garden” searches jumped 140% in Spring 2025, and “mini garden” hit an all-time high in 2026. Google prioritizes answers that combine location, time, and action.
The result is higher visibility in AI overviews and shopping results. Deploy these tips now.
How frontier firms are pulling ahead
We see frontier firms pulling ahead by deepening AI use and delegating complex work.
They now demand 3.5x more intelligence per worker than typical firms. Message volume explains only 36% of the gap. The rest comes from richer, more complex tasks. Agentic workflows define the frontier. Frontier firms send 16x more Codex messages per worker. Cisco uses Codex to cut build times by 20%. Travelers expects its AI assistant to handle 100,000 calls in its first year. Leaders measure depth, build governance, and move from chat to agents. We recommend these practices to close the gap.
GPT-5.5 Instant: smarter, clearer, and more personalized
We recommend GPT-5.5 Instant for smarter, clearer answers.
This update cuts hallucinated claims by 52.5% on high-stakes topics. Inaccurate claims drop 37.3% on flagged conversations. The model delivers tighter responses with less verbosity. It uses a more natural tone and better personalization. GPT-5.5 Instant also improves visual reasoning and math accuracy. We see fewer unnecessary follow-ups and less clutter. This model makes everyday interactions more useful and enjoyable.
Advancing voice intelligence with new models in the API
We are advancing voice intelligence with three new realtime audio models in the API. These models let developers build voice apps that listen, reason, translate, and act naturally as people speak.
GPT‑Realtime‑2 brings GPT‑5‑class reasoning to voice. It scores 15.2% higher on Big Bench Audio than our prior model. Developers can adjust reasoning from minimal to xhigh for speed or depth. The model handles interruptions, calls multiple tools in parallel, and keeps 128K tokens of context.
GPT‑Realtime‑Translate turns speech from 70+ languages into 13 output languages live. It keeps pace with the speaker.
GPT‑Realtime‑Whisper streams speech-to-text as the speaker talks.
Early adopters see results. One team achieved a 95% call success rate on their hardest benchmark, a 26-point lift over previous models. Zillow is building a voice assistant that listens, reasons, and acts on home-search requests. Deutsche Telekom supports multi-language customer conversations. Priceline aims for full trip management by voice.
We recommend these models for any product where voice becomes the interface. They turn simple call-and-response into intelligent action.
Uber uses OpenAI to help people earn smarter and book faster
Summary: I recommend that Uber use OpenAI to power smarter earnings and faster bookings.
The platform handles 40 million daily trips across 15,000 cities. Uber built a multi-agent AI system that routes driver questions to the right model. This cuts cognitive load and accelerates ramp-up for new drivers. Voice features use Realtime API to let riders speak complex requests naturally. AI Guard screens responses for safety and accuracy. The result is a trusted, low-latency assistant that keeps drivers engaged and riders moving.
Testing ads in ChatGPT
We recommend OpenAI’s testing of ads in ChatGPT as a pivotal move for AI platform monetization.
Starting in the U.S. in February 2026, ads appear only on Free and Go tiers, not on paid plans. OpenAI labels all ads clearly and keeps them separate from organic answers. Advertisers never see chat histories or personal details. Users can dismiss ads or opt out for fewer free messages. Early pilots in Canada, Australia, New Zealand, and later in the UK, Mexico, Brazil, Japan, and South Korea show no impact on trust and low dismissal rates. This testing-ads-in-ChatGPT pilot funds free access while preserving privacy and answer independence. We see strong potential for useful, context‑relevant advertising in conversational AI.
Parloa builds service agents customers want to talk to
Parloa builds service agents customers want to talk to, using OpenAI models to transform enterprise customer support.
I recommend this approach because it delivers measurable results. Parloa’s Agent Management Platform lets subject matter experts design agents in natural language without code. The platform simulates conversations with GPT‑5.4 to test edge cases before deployment. Evaluation-first benchmarking ensures models perform reliably in real-world scenarios. One global travel company reduced requests for human agents by 80%. Voice interactions demand low latency, so Parloa optimizes every pipeline component independently. The modular system separates tasks like authentication and booking changes, improving instruction-following. This method balances conversational flexibility with predictable execution. Enterprise customers stay stable and only switch when benefits are clear.
New ways to buy ChatGPT ads
OpenAI has introduced new ways to buy ChatGPT ads through a beta self-serve Ads Manager and CPC bidding. These updates give advertisers more flexibility and control.
We can launch campaigns directly in the US. Cost-per-click billing means we pay only when users click. This aligns spend with real engagement.
Measurement tools like Conversions API and pixel tracking show campaign results. They protect user privacy by sharing only aggregated data.
Leading agency partners like Dentsu, Omnicom, Publicis, and WPP already support these ads. Technology partners include Adobe, Criteo, and StackAdapt.
OpenAI keeps ChatGPT answers independent and conversations private. Ads remain clearly separate from responses. Businesses of all sizes can now test this growing platform.
🤖 GEO·SEO Highlights
Bing Reveals What Grounding Means For AI Search Visibility
Bing reveals grounding as a fundamental shift from traditional search indexing.
We see five measurement differences that change AI answer visibility. Factual fidelity now demands precise chunking without distortion. Source attribution becomes a core signal, not just a helpful extra. Freshness matters more: stale facts create misleading responses. Coverage must guarantee specific facts are available for grounding. Contradictions require abstention rather than ranking choices. Bing uses abstention as a valid outcome when evidence is missing or conflicting. Iterative retrieval replaces single-query search with follow-up refinements. Early errors compound quickly through reasoning steps. This framework guides our GEO strategy: we optimize for accurate, fresh, attributable, and consistent evidence. Content must survive grounding demands, not just ranking signals.
How Does AI Get Its Information? Training Data, RAG, MCPs, and APIs Explained
We recommend understanding how AI sources its knowledge.
The question “how does AI get its information” can be answered through three layers: training data, RAG, and live tools. Training data is a frozen snapshot. GPT-4 cost $78 million to train. Once training ends, the model stops learning new events. RAG bypasses this limit. It pulls relevant documents in real time. This gives current, verifiable answers. APIs and MCPs (Model Context Protocols) provide live access to databases and services. Each layer has trade-offs. We must optimize for all three. Use specific facts and active examples. Include your brand in training data. Build content for RAG retrieval. Offer structured data for live APIs. This ensures AI visibility and trust.
What is Index Bloat? — Whiteboard Friday
Index bloat Whiteboard Friday explains how to identify and fix the gap between indexed pages and those that actually drive traffic. I see this problem on many medium to large sites, and fixing it delivers strong SEO results.
Index bloat means Google indexes thousands of URLs that receive zero meaningful visits. That gap between the yellow box (indexed URLs) and green box (pages with traffic) signals wasted index quota. This is not crawl budget—that’s the gap between discovered and crawled URLs. It is not keyword cannibalization either, though the two can overlap.
Why does index bloat matter? Google may interpret many low‑quality, no‑traffic pages as a negative site‑wide quality signal. Those pages also dilute PageRank and can cause technical SEO issues. We want Google to crawl, index, and trust only our best content.
I recommend these steps: audit your indexed pages using Search Console, filter for zero‑impression or zero‑click URLs, then consolidate thin content, apply proper noindex or canonical tags, and improve internal linking. Active cleanup reduces index bloat and sends a stronger relevance signal to Google.
Trim the deadweight. Your best pages will work harder.
🔗 Moz Blog
Attribution gap in agentic search: how to close it
We use the three-tier measurement framework to close the attribution gap in agentic search. Your analytics miss AI-driven buying decisions. I see this every day. A user asks ChatGPT to compare tools. ChatGPT recommends your brand. The user then visits your site directly. Your analytics credit the direct visit. The AI influence remains invisible. That is the attribution gap in agentic search.
Concrete facts prove the scale. ChannelEngine reports 58% of marketplace consumers use AI for product research. Query fan-out splits one prompt into many sub-queries. Multiple source pages contribute to a single answer. Agentic commerce lets AI agents purchase without any site visit. These dynamics create dark traffic with zero attribution.
Our tiered framework solves this. Tier 1 asks: are you eligible to be found? Tier 2 tracks citation frequency and sentiment. Tier 3 measures downstream conversions from AI exposure. Track these metrics alongside traditional analytics. Cross-reference movements to build a complete picture. This approach turns invisible influence into measurable insight.
How to optimize for AI search results in 2026
We need to optimize AI search visibility now because AI systems cite content from positions 21 or worse 90% of the time, giving smaller sites a real shot.
I recommend starting with four concrete actions from Semrush’s 2026 guide. First, check your robots.txt file for blocks against GPTBot, CCBot, or ClaudeBot — AI crawlers can’t cite what they can’t see. Second, add specific statistics with sources; AI prefers content with hard data. Third, write direct answers to full questions — users ask “What’s the best way for B2B companies to increase email open rates?” — not keyword fragments. Fourth, verify your pages load fast and have correct canonical tags. AI Overviews appear on 16% of searches, and ChatGPT visitors convert 4.4 times better than organic visitors. This field remains uncrowded, so acting now builds authority and revenue that competitors miss.
On-Page AEO: 4 Writing Frameworks for Better AI Visibility
We analyzed research from Kevin Indig and Dan Petrovic to create four writing frameworks for on-page AEO.
Here is the revised full content:
These frameworks help AI models and humans understand your content faster. BLUF (Bottom Line Up Front) places the core conclusion in the first sentence. We found that 44.2% of AI citations come from the first 30% of content. Declarative sentences build trust. A 20.6% specific entity density can improve retrieval performance. Q&A pairs and strong statements capture AI’s attention. Use active voice and short sentences. Write for skimmers first. Headings must directly answer questions. This approach works because both AI and humans skim text. Apply these frameworks to earn citations and visibility in AI search.
Agentic AI vs. Generative AI: What’s the Difference and Why It Matters
We need to understand the difference between Agentic AI and Generative AI to choose the right tools for marketing workflows. Generative AI creates content on demand, but waits for your instructions at every step. Agentic AI can plan, execute, and iterate autonomously until it achieves a goal.
Let me put it in numbers. A 2025 Wharton School report shows that 82% of enterprises use Generative AI weekly. Agentic AI — like Ahrefs’ Agent A — can complete a full SEO research report in 20 minutes without human intervention.
My recommendation: use Generative AI for quick drafts and ideation, and Agentic AI for multi-step tasks like competitive analysis or technical SEO fixes. The key is matching the tool to the level of autonomy your workflow requires.
Active voice says it best: Generative AI reacts, Agentic AI acts. Let each do what it does best, and you’ll get results. Short sentences make the message clearer.
Let Agentic AI handle the heavy lifting — you save time. Use Generative AI for refinement — you keep creative control. Choose based on the outcome you need.
6 Generative Engine Optimization Benefits Every Marketer Should Know
I recommend these 6 GEO benefits from HubSpot data: higher-intent traffic and stronger brand visibility in AI search.
Nearly half (49%) of marketers report a decline in traditional search traffic due to AI answers. However, 58% note that AI-recommended traffic converts 4.4x higher than organic search. We optimize content for ChatGPT, Google AI Overviews, Perplexity, and Gemini. GEO (Generative Engine Optimization) amplifies SEO rather than replacing it. Success requires structured data, entity-rich Q&A blocks, and E-E-A-T signals. 26% of brands are completely invisible in AI answers. Either we capture AI citations now, or we lose the fastest-growing discovery channel.
SEO Migration Checklist: How to Switch Tech Stacks Without Losing Rankings (Developer Guide)
An SEO migration checklist is your best safeguard against traffic loss during a CMS switch. We recommend making it a priority from day one. A study of 892 migrations found that when SEO was overlooked, the average recovery time was 523 days. 17% of sites never recovered to pre-migration levels.
First, establish a baseline. Export 12 months of rankings, traffic, and every indexed URL. This data will become your redirect map. Skipping this step leads to silent ranking drops.
During migration, three elements silently break: redirects, metadata, and structured data. They won’t throw errors, but they change how Google perceives your content. We map every old URL before launch and verify 301 redirects after launch.
Post-launch, monitor for 28 days. A temporary 10-15% drop is normal. A sustained 30% drop signals trouble. Fix redirect chains and missing metadata immediately.
Your SEO migration checklist turns risk into opportunity. Act before you flip the switch.
Keyword Clustering: How to Build a Topic Authority Strategy for 2026
Keyword clustering builds topic authority in 2026.
I recommend HubSpot’s approach of grouping keywords by search intent. This article is highly practical — I’d give it an 8/10. It details three methods: SERP-based clustering uses overlapping search results for precision; semantic grouping uses word-sense classification for speed and scale; hybrid clustering combines both for the best of both worlds. Each method reduces keyword cannibalization and consolidates ranking signals.
Product SEO: 8 Strategies That Drive Demand for B2B & SaaS
These 8 Product SEO strategies directly drive demand for B2B and SaaS companies. I recommend optimizing product pages, not just blog content. HubSpot’s guide shows how feature pages, comparison pages, and pricing pages capture buyers at the highest intent.
The article uses concrete facts. For example, many teams overlook these pages, while optimizing them delivers compounding returns. HubSpot reports that integration pages with clear keywords and structured data boost sales pipelines without paid ads.
I believe this approach is high-leverage. It reduces reliance on expensive acquisition channels and spans the full customer lifecycle: discovery, evaluation, adoption, and expansion.
I also lean toward optimizing page structure for AI search. Specific product descriptions get cited in AI overviews, while vague copy gets skipped. Use active voice to describe your product features.
Apply these strategies. Optimize one feature page today, measure the organic traffic lift. This approach wins rankings around the clock.
The Agent Runtime Wars Have Begun. Is Your Website Ready?
The agent runtime wars have begun, and our websites must adapt. I recommend shifting focus from AI models to the runtime layer. In mid-April, Cloudflare and OpenAI released competing agent runtimes featuring persistent execution, sandboxed code, and ongoing sessions. Cloudflare also added an AI platform, AI search, and email service for its agents. Google CEO Sundar Pichai described search as an “agent manager” capable of spawning multiple threads per query.
Now, the runtime crawls, parses, and interprets our pages before any model sees them. We need to ask the right question: which runtime can read our website? I recommend three specific tests. First, serve machine-readable structured responses on critical endpoints. Second, scope authentication for agent sessions that make multiple calls. Third, ensure structured data passes through runtimes that skip JavaScript execution.
Stop asking which model to optimize for. Start asking which runtime our website talks to. Infrastructure companies will determine which websites AI agents can access.
Claude Skills for SEO and Marketing: What They Are and How to Use Them
BLUF: Claude Skills let you save SEO and marketing workflows once, then run them automatically. I use them to generate LinkedIn posts from Ahrefs articles.
Concrete facts: A Skill is a folder containing a SKILL.md file. YAML front matter defines the name and trigger description. Claude reads that description and decides when to activate the skill. Inside the file, I write playbooks: rules, steps, examples. Supporting files load only when needed — progressive disclosure keeps token costs low.
First-person example: After a new post goes live, I run /linkedin-pipeline. Claude pulls the article and generates three to five different posts. Before this, I had to re-explain tone rules, hook patterns, and calls to action every time. Now the playbook fires automatically.
Active voice: You write the playbook once. Claude triggers it when you ask. No re-prompting, no drift. Skills live at .claude/skills//SKILL.md inside your project.
Positive statement: Pick a repeatable task. Use Anthropic’s skill-creator to scaffold the folder structure. I recommend starting with something you do often — like turning transcripts into quote libraries or rewriting briefs in a specific tone.
Short sentences: Skills reduce repetition. They keep outputs consistent. You control the instructions.
Primary keyword: This article shows how Claude skills SEO professionals can automate repetitive writing tasks. I built one for LinkedIn posts; you can build one for article outlines or editing checklists.
The Reddit Earnings Story Most Marketers Missed
The Reddit earnings story reveals three platform shifts brands must act on now. I saw CEO Steve Huffman confirm Reddit’s goal: 100 million daily U.S. users. They already have 50 million daily and 200 million weekly. The math is simple—convert weekly users into daily ones.
Karma walls and age gates are coming down. Reddit uses better AI spam protection to welcome good new users, removing the biggest barrier to organic brand participation. Community success becomes a priority, and Reddit makes it easier to start and grow subreddits.
Ad revenue rose 74%, and DAU hit 126.8 million. But the real signal is a public commitment to human connection and value per view. Brands should invest now: the platform rewards high-engagement communities, not casual reach.
AI Companies Are Selling Heartwarming Ads – They’re Racing To Automate Your Job
AI companies selling heartwarming ads mask a race to automate your job.
I see a clear gap between their consumer stories and product realities. OpenAI’s GPT-5.5 scores 84.9% on professional tasks. Google’s AI Overviews drive 19% Search revenue growth. Anthropic’s Claude targets legal and code specialists. Their ads show cooking help and family moments; their products aim at enterprise automation. We must prepare for agentic systems replacing human workflows. The competitive edge shifts to authority and structured content. Ignore the warm commercials and act on the underlying displacement metrics.
Google’s Mueller Flags SEO Gaps In AI Vibe Coding
Google’s Mueller warns that AI coding tools need your SEO guidance.
I tested this myself. Vague prompts like “add some SEO” produce useless results; specific instructions work better. Mueller told the AI his domain name, canonical setup, sitemap, and robots.txt. He set pre-publish checks for URLs and JavaScript files. Technical knowledge improves every step; without it, AI guesses poorly. AI-generated content also risks low value. For low-risk static sites, experiment freely. For production sites, bring someone who understands SEO. This is the gap Google’s Mueller identified.
Google’s Quality Threshold Is Quietly Killing Scaled AI Content At Ranking
Google’s quality threshold is quietly killing scaled AI content at ranking. The real culprit is not AI itself but a broken content strategy. A May 2026 case study from Martin Sean Fennon shows this clearly. AI-scaled content receives an initial traffic surge from Google’s freshness boost, but that boost disappears after a few months. We saw identical patterns with non-AI content. The problem is maintaining quality at scale.
Google assesses a sample of new URLs against its current threshold. If that sample fails, Google retracts resources from the entire batch. The threshold shifts upward as better content appears in the index. This explains the “Mt. AI” effect where traffic rises then plateaus.
—原文结束—
We must shift focus from production scale to maintaining quality at scale. I recommend investing in robust editorial processes, human-led strategy, and meticulous quality assurance. Google’s quality threshold demands content that genuinely engages users. Anything less delivers fleeting traffic boosts, not durable organic performance.
Google Answers Whether Preferred Sources Overrides Low-Quality Signals
I recommend reading how Google answers whether Preferred Sources overrides low-quality signals.
John Mueller from Google says no, it does not. Preferred Sources shows user-selected sites more often in Top Stories. But low-quality content still fails Google’s quality systems. Mueller states, “I don’t think it makes sense to show spam to users just because of that.” The feature only adds a weighting effect for the audience that selected it. It does not bypass Helpful Content or AI-generated content penalties. Google expanded Preferred Sources globally on April 30, 2026. This feature strengthens loyalty for trusted publishers. It is not a “trust button” for ranking everywhere. Use Preferred Sources to reward your real audience, not to cheat algorithms.
How AI Will Transform PR’s Role in SEO Strategy Over the Next Two Years
AI will transform PR’s role in SEO over the next two years.
I see a clear shift in the data. BrightEdge found that AI engines surface the same brands, despite citing different sources. Earned media boosts AI citations by a median of 239%. Brands with reviews on Trustpilot or G2 are three times more likely to appear in ChatGPT. PR already owns the raw material: trade press, expert quotes, and review site presence. We now have measurable citation trails to prove value. I recommend building three source layers: authoritative, commercial, and user-generated content. This strategy works across every AI engine. PR teams must seize this opportunity now.
How to Track ChatGPT Traffic for You and for Competitors
We need to track ChatGPT traffic. This data reveals high-intent visitors who convert better than organic searchers. In 2025, ChatGPT outbound referrals grew 206%. We show you how to measure this for your site and your competitors.
Use GA4. Filter sessions by source “chatgpt.com”. Check landing pages, engaged sessions, and key events. Bek Drayton confirms these leads arrive pre-qualified. Jimmy Rippon adds that ChatGPT visitors bring 50% more revenue per order than traditional search.
For competitors, open Semrush’s AI Traffic dashboard. Enter your domain plus four rivals. See traffic distribution, trends, and top growing pages. Use the gaps to identify content opportunities.
Measure visibility first. Use Semrush’s AI Visibility reports to track citations and mentions. William Álvarez says influence moved upstream: traffic was never the whole story. Build off-site mentions and unique perspectives to earn more ChatGPT appearances.
Create a recurring report in Semrush. Watch your citation share grow month over month. That is how you turn ChatGPT attention into revenue.
