seolist.org Miniblog

How to Spot an SEO Agency That Delivers, Not Just Talks

Anyone hiring an SEO agency should treat the sales pitch like a blind date: charm is cheap, and the bill arrives later.

The practical test is simple. Don’t ask for miracle rankings; ask how success will be measured. Google’s Search Essentials makes clear that nobody, not even Google, can promise a fixed spot in results. Useful measures are qualified traffic, commercially relevant keywords, assisted conversions and a rise in branded searches. Page-one boasts on their own are often window dressing. One outfit even hit a 60-day ranking promise by getting a company to rank for its own brand name, which is rather like claiming to have discovered your own front door.

Competence shows up in business understanding. A serious agency should know your competitors, ask about customer lifetime value and explain why some high-volume terms are useless bait. If a 45-minute call passes without questions about what you sell, who buys it and what margin you make, you’re not being advised; you’re being put through the mangle.

Ask for a domain-specific audit, even a small one, not a glossy deck. Ask to see redacted monthly work: reports, briefs, technical checklists, outreach sheets. Also pin down who will actually handle the account.

For competitive B2B SaaS, effective retainers usually start above $5,000 a month; past $15,000, expect full content, technical and outreach work. SEO is a long game. Better less of it than cheap busywork.

Posted on 3 July 2026

Meta Descriptions: Optional, Awkward, Still Worth Doing

Meta descriptions are in the odd position of being both nonessential and still useful, which is very internet. Google’s John Mueller recently clarified on Reddit that there is no ranking penalty for leaving them out, and no requirement to write them at all. That has broadly been true for years, despite recurring SEO despair whenever Google swaps in its own snippet.

Google’s documentation says much the same: if writing descriptions for every page is too much, prioritise the important URLs, especially the homepage and popular pages.

That matters because a handcrafted description can still do jobs Google’s autogenerated snippet may fumble. On product pages, crucial details such as price, manufacturer, age, model, or reviews often live in separate parts of the page. A concise meta description can gather those fragments into one clear summary. The same goes for articles or blog posts, where author names, publication dates, and bylines may deserve prominence.

There is also a branding angle. Search snippets generated by Google are not always elegant, and businesses may prefer to control how key pages present themselves in results.

Mueller also made a more interesting point: writing a description forces clarity. Distilling a page into one or two sentences can reveal what it is actually about, and whether that matches the intended audience and topic focus.

So: optional, yes. Pointless, no. Especially on pages that matter. Posted on 30 June 2026

Google’s AI Answers Often Use a Brand’s Pitch, Then Recommend a Rival

Google’s AI Overviews have a sly habit: they will borrow from a company’s own best-of page, then hand the bouquet to somebody else.

A new analysis by Lily Ray tracked 100 B2B searches for best [category] software on April 15, May 15, and June 8. Eighty of those queries produced an AI Overview. Across them, Google cited self-promotional listicles 323 times. Yet in 224 instances, the brand supplying the cited page was left out of the actual recommendations. That works out to 69% of cases.

The pattern showed up all over B2B software. For best LMS for selling courses, Google used Oasis LMS as a source, then recommended Kajabi, Thinkific, LearnWorlds, and Teachable, all named in Oasis LMS’s own article. Similar mismatches appeared in help desk, task management, survey, CRM, and SEO software.

The winners were usually stronger incumbents: brands with broader third-party mentions, category leadership, and better link profiles.

Ray also found that sites leaning hard on self-ranked best pages, AI-generated articles, comparison content, and GEO-focused publishing often saw organic traffic decline starting around Jan. 20, with losses deepening during Google’s May 2026 core update. Earlier work had found 30% to 50% visibility drops for some SaaS and B2B brands.

Meanwhile, Google’s citations increasingly favored outside sources such as Forbes, Reddit, and YouTube, with Reddit rising sharply. The lesson is plain: being cited is not the same as being chosen.

Posted on 20 June 2026

When a Quiet Export Rule Becomes an On-Off Switch for AI

By late Friday, Anthropic had taken Fable 5 and Mythos 5 offline for everyone, not because the models had run amok, but because the U.S. Commerce Department sent a nonpublic enforcement letter invoking an obscure export-control rule. The order barred non-Americans, including Anthropic staff, from accessing the systems over an unspecified national-security concern. No court process appeared necessary.

Anthropic believes the trigger was a reported guardrail bypass, though the letter itself gave no details. That vagueness is the real story. A government agency demonstrated that it can, with remarkable speed, force a major U.S. software company to disable core products.

Weekend reporting complicated the official rationale further. Accounts from people familiar with the matter suggested the directive may have arisen less from a technical failure than from friction between Anthropic and the administration.

Katie Moussouris of Luta Security, after reviewing a private paper Anthropic shared with her, said the alleged bypass described by researchers at Amazon amounted to little more than reframing a request from identifying security flaws in code to fixing them. In practice, she argued, the capability is substantially the same, and trying to stamp it out would blunt defensive uses. She and many other security experts urged the government to reverse course, warning that removing advanced cyber capabilities from U.S. defenders is dangerous.

Justin Hendrix noted the broader damage: allies may now question whether American AI systems can be relied on for critical work if official intervention can arrive this abruptly.

Posted on 16 June 2026

When Google Stops Sending You Away

Google search used to be a vending machine: type a need, get links, wander off. Now it’s more like being trapped in an airport lounge with a robot concierge reading at you.

Analysis of about 846,000 anonymized U.S. Google sessions with Eric Van Buskirk of Clickstream Solutions found that when AI Overviews appear, classic intent signals stop predicting SERP behavior. Without an AIO, only 12% of navigational users are still on the results page after 21 seconds, versus 32% of local searchers. With an AIO, all five major intents — informational, local, navigational, transactional, and video — bunch together between 41.9% and 48.5%. Average SERP sessions become nearly four times longer.

That’s the real shift: search becomes a reading session. The old model sorted behavior by intent; AIOs flatten it. Direct answers keep people on Google longer because users now consume context before deciding whether to click.

For brands, that makes the second impression crucial: the back-scroll after users pass a listing once. Product pages need schema covering aggregateRating, reviews, offers, availability, plus strong review volume and multiple images. Category pages need ItemList schema, visible facets, and real depth. Blog posts need fresh datePublished or dateModified signals and named authors tied to sameAs URLs.

The content strategy built over the last three years still stands. What changes is the battleground: not what to publish, but how each result competes beside an AI answer.

Posted on 9 June 2026

Google Gives Publishers a Discover Front Door With Search Profiles

Google is giving publishers a new front porch inside Discover: Search profiles, now rolling out in the U.S. after months of testing and recent tweaks such as shortnames.

The feature creates a dedicated landing page for eligible publishers and creators. On one screen, users can see a large header image, follow the source, and browse its newest articles, videos, and social posts. The larger goal is obvious: keep audiences connected to trusted outlets across platforms while giving Discover another reason to act like the internet’s neighborhood newsstand.

Search profiles are being offered first to publishers and creators with sizable audiences on at least one major social or video platform. The current minimums are 300,000 followers on TikTok, or 100,000 on YouTube, Instagram, or X.

Eligible accounts can claim and customize a profile with an avatar, bio, website, and links to social and video channels. In some cases, claiming a profile may also create a Knowledge Panel. If a Knowledge Panel already exists, Google says it can be upgraded with a refreshed avatar, newer content, and a direct profile link.

For publishers, this is another Google visibility play: a cleaner brand hub, more chances to gain followers, and one more lane for surfacing recent work. The unresolved question is whether that added exposure will matter enough as AI keeps changing how people use search.

Posted on 8 June 2026

Google Trials Healthcare Ads in AI Mode With Tight Early Limits

Google has started a limited U.S. trial of healthcare advertising inside AI Mode, its AI-led search experience, a small experiment with rather large implications.

The test covers English-language queries only and, for now, is restricted to healthcare advertisers. Google Ads Liaison Ginny Marvin confirmed the rollout after industry chatter about healthcare promotions appearing in AI-generated search results.

Several campaign types can be eligible in this test: Performance Max, AI Max with search term matching, Shopping campaigns, and broad match campaigns. Those same formats can also run in AI Overviews.

That matters because healthcare is among Google’s most tightly controlled ad categories. If adverts can be made to work here, within a space where accuracy, trust and regulation are not optional extras, it offers an early clue to how Google may earn revenue from AI search more broadly.

There are limits. The opening phase excludes creatives using pinned assets or text disclaimers, which reduces the number of healthcare ads that can participate.

So this is not yet a grand parade of AI-age medicine marketing; it is more like a carefully supervised lab test. Still, advertisers will be watching for any widening of access: more healthcare brands, more ad formats, and perhaps eventually other regulated sectors. The deeper question is how Google intends to fit monetization into AI search without unsettling user confidence.

Posted on 3 June 2026

How to Get Your Ads Inside Google’s AI Overviews Without Losing Your Mind

Google Search is turning into that friend who answers their own question before you can speak, which means advertisers now have fewer chances to win a click and more need to show up inside AI Overviews.

Google’s best bets for that placement are Shopping, Performance Max, and AI Max for Search. Translation: the campaign types many marketers side-eye because automation feels like letting a Roomba run your budget. But these formats are built for looser matching, where intent matters more than exact keywords.

Shopping lives or dies on feed quality: strong images, detailed titles and descriptions, and even optional attributes. That feed helps Google decide when a product belongs in AI-generated carousels tied to high-intent searches.

Performance Max leans on page content, product data, and audience signals. Turning on Final URL expansion gives Google more chances to match ads to relevant searches using landing-page context. AI Max for Search starts with existing keywords, then stretches beyond them with broad match, smart bidding, Final URL expansion, and asset optimization.

To improve your odds, use varied creative in multiple formats, write like a helpful human instead of a coupon siren, answer the who/what/how/why directly, tighten schema markup, add credible links, feed the system strong first-party audience signals, and keep monitoring search terms, landing pages, exclusions, negatives, and auto-created assets for brand safety and profit.

Posted on 2 June 2026

Small Businesses Are Disappearing in AI Search for Three Fixable Reasons

Search has changed shape, and many small businesses are still feeding the bones of the old machine. The old bargain was simple: rank, win the click, hope for a sale. But a Pew Research Center study of real browsing behavior shows that when Google displays an AI summary, users click a standard result only 8% of the time; without that summary, the rate is 15%. Similarweb reports zero-click searches rising from 56% to 69% in roughly a year.

Three habits now make firms hard to find.

First, pages built for keyword density fail when AI systems hunt for compact answers. These tools often lift a concise passage and present it directly. If the useful response is hidden deep in the page, another business with a clearer opening may be chosen instead. A short, stand-alone answer near the top now matters.

Second, a business that exists only on its own website is weaker in AI search. Pew found more than 88% of AI summaries cite at least three sources. Models draw from directories, reviews, press coverage, forums, industry sites and Wikipedia. Consistent mentions across the web build trust signals.

Third, businesses still chasing short keywords miss how people now search. Instead of plumber Birmingham, users ask long, detailed questions. Pew found searches of 10 words or more trigger AI summaries about 53% of the time, versus 8% for shorter queries.

Clarity, not gaming, is becoming the survival skill. Posted on 1 June 2026

Google Search Has Become a Browse Experience

Google search now has two personalities, and they don’t even text each other back.

A new analysis of about 846,000 anonymized U.S. Google sessions from February and March 2026 shows that AI Overviews and AI Mode trigger opposite user behavior. AI Mode pushes people forward fast; AI Overviews make them linger, compare, and often scroll back up, more like browsing Netflix than picking a link and leaving.

The study, conducted by Eric Van Buskirk of Clickstream Solutions using clickstream data provided by Surfer SEO, is the fifth public look in 12 months at how Google users react to AI search features. Earlier work included a 70-user UX study in May 2025 and a 250-session AI Mode study in October 2025. This one sacrifices intimate screen-recording drama for scale, drawing from tens of thousands of search users, far beyond older mouse-tracking research that usually covered only dozens of people or a few thousand tasks.

Four shifts stand out: AI Overviews produce longer result-page evaluation, more reverse scrolling, weaker behavioral differences across search types, and fewer shortcuts on brand queries.

The practical implication is painfully modern: title tags and meta descriptions matter differently now. If users are looping, reassessing, and delaying clicks, search snippets have to win not just the first glance, but the second and third existential crisis too.

Posted on 31 May 2026

Google Pushes Trusted Sources Deeper Into AI Search

Google is giving its AI search products a little more main-character energy, with new tools meant to surface trusted reporting and firsthand takes faster.

The biggest change: Preferred Sources now appear inside both AI Mode and AI Overviews. If you’ve already picked publications or sites you trust, links from those outlets will be marked with a Preferred Sources label inside AI-generated answers, making them easier to spot. The feature is rolling out globally in every language. Google says people have already chosen more than 345,000 unique sources, and users are twice as likely to click a Preferred Source.

Another addition is a new perspectives carousel in AI Mode and AI Overviews. Google plans to show it when a search suggests someone wants to go deeper on an evolving topic. The carousel can feature timely articles, including content from a user’s Preferred Sources, and is designed to give fresher reporting more visibility across a wider set of queries.

Searchers may also see a related carousel built around perspectives pulled from online discussions, forums, and social media, which feels very internet in 2026.

Google is also widening use of its Highly Cited label across Search, beyond AI features. It will now appear on more article links, and Google will also note when a story directly cites a Highly Cited source, helping users identify the original reporting other coverage relies on.

Posted on 29 May 2026

When AI Enthusiasm Outruns Management

A familiar pattern is returning to technology: exuberance first, operational reality later. Yet the AI cycle is stranger than earlier transitions such as cloud computing. Companies are posting record revenue while dismissing workers at a historic pace.

One diagnosis came from Box founder Aaron Levie, who argued that CEOs are especially vulnerable to AI grandiosity because they are far from the final, error-prone stages where value is actually created. A prototype that drafts a contract or writes code can look transformative from the executive floor; less visible are bug hunts, hallucinated libraries, bespoke legal terms, and the tedious review required before output becomes usable. Levie is no skeptic: he promotes AI heavily, writes about agent-oriented software, and invests in AI startups. His advice is immersion—use AI constantly until its strengths and limits become equally clear.

That lesson is not widespread. Layoffs.fyi reports 115,430 tech workers cut by 152 companies in the first five months of 2026, versus 124,636 across 275 companies in all of 2025. Many firms cite AI, though critics see AI washing.

ClickUp CEO Zeb Evans said 22% of staff were let go after deploying about 3,000 internal AI agents, aiming for a 100x organization built around humans supervising agents.

The evidence remains sobering: Berkeley found no robust link between AI adoption and aggregate productivity; NBER found gains smaller than perceived; MIT estimates 80%–95% minimally sufficient performance on most text tasks by 2029, with true human outperformance later. HBR suggests the bottleneck simply migrates upward—to executives themselves.

Posted on 28 May 2026

Google Search Console’s Link Report Reverted to Older Data After Outage

For a few days, Google Search Console’s links report became a faulty mirror. Beginning Thursday, some site owners opened it to find no links at all; others saw losses nearing 90% compared with the week before. The report had not become a record of reality, only of a break in the machinery.

Google acknowledged the problem. John Mueller first said the company would investigate, though the long weekend might slow the response. By Saturday, the missing links seemed to have returned, but the restoration was only temporary: Google had rolled the report back to data from the prior week while engineers worked on the underlying issue.

So the links report is currently visible, but not current. What it shows now is older data from weeks ago, not a fresh account of how Google is seeing links today.

That distinction matters. Anyone using the report for client updates or stakeholder summaries should treat recent pulls with care, especially data exported on Thursday, when the bug was at its most misleading. In some cases the report displayed zero links; in others, more than 85% appeared to vanish.

Google is still fixing the problem. Until that work is done, the report remains a useful tool only if users remember that its numbers belong to an earlier moment.

Posted on 27 May 2026

Google’s AI Search Advice Sounds Reassuring. That’s Why It Deserves Suspicion.

Google’s latest AI-search advice has big It’s Fine energy, and that’s exactly the problem. The company says optimizing for generative search is still SEO, while dismissing tactics like llms.txt, passage-level rewriting, and chunking. But that framing feels less like clarity and more like brand management.

The skepticism isn’t random. In 2024, leaked Content Warehouse documents exposed a gap between Google’s public guidance and its internal ranking systems, including signals it had downplayed. That history makes any tidy mythbusting document hard to take at face value.

The bigger issue is practical: AI search now spans Google, ChatGPT, Perplexity, Claude, Copilot, Gemini, and specialist agents. Traditional SEO still matters—crawlability, structure, page experience, distinctive content—but it no longer covers the full job. AI visibility increasingly depends on passage retrieval, vector similarity, synthesis pipelines, and off-site presence across places like Wikipedia, Reddit, third-party media, and licensed data sources.

Microsoft has been notably more explicit. Teams led by Krishna Madhavan and Jordi Ribas have described a shift from ranking pages to retrieving verifiable, groundable facts, shipped AI Performance in Bing Webmaster Tools, and acknowledged emerging GEO work.

Google may not use llms.txt, but Anthropic does. Google says chunking is unnecessary; Bing and Google’s own research on passage indexing suggest otherwise. In other words: useful advice, incomplete map. Posted on 25 May 2026

Google Tells Sites to Stop Chasing AI Tricks and Start Making Better Pages

Google has now handed website owners a tidy little handbook for the age of machine-made answers, publishing a new help document on May 15, 2026 that explains how to prepare pages for Search features such as AI Mode and AI Overviews.

The central message is almost comically simple: the old rules still matter. Traditional SEO remains the backbone for visibility in generative search. Google advises publishers to make material that is useful, dependable, people-first, and distinctly their own rather than churned-out commodity copy. A clear viewpoint, sensible organization, strong images and video, and restraint against stuffing pages with too much are all part of the recipe.

The guide also leans on technical discipline. Sites should meet Search requirements, support crawling, keep page experience solid, reduce duplication, and maintain a clean technical structure. Semantic HTML should be written for human readability, while JavaScript should follow existing best practices. It also flags local business and ecommerce information as worth tightening up.

Perhaps the juiciest section is the mythbusting. Google says there is no need for LLMS.txt files, special markup, content chunking, rewrites aimed at AI systems, or fake mentions. It also warns against obsessing over structured data.

The document further nods to agentic experiences and serves as a single place to gather guidance Google has previously spread across blogs, videos, events, and other materials.

Posted on 24 May 2026

Google’s New Search Box Wants to Be the Internet Instead of a Map to It

For decades, Google’s blank little rectangle was the front desk of civilization: type a few words, get a stack of blue links, pretend you’re in control. At Google I/O in Mountain View, California, on May 19, 2026, Alphabet CEO Sundar Pichai said that box is now being rebuilt around AI—the biggest search overhaul in more than 25 years.

Google says Search will run through Gemini 3.5 Flash, not just “AI Mode.” Results won’t merely point outward; they’ll open with an AI-written summary and a back-and-forth interface much like ChatGPT. Users can also search with images, files, videos, or even Chrome tabs. Google says its suggestions will go “beyond autocomplete.”

The convenience comes with collateral damage. TechCrunch declared, “Google Search as you know it is over.” The Verge’s Jay Peters wrote, “It’s not hard to imagine a future where, someday, Google just makes everything happen in one universal search box,” while warning that if Google does everything, much of the web beneath it could buckle. Publishers had already called earlier AI search changes “catastrophic.” Small businesses saw traffic drop. Lily Ray of Amsive warned the shift would be “going to have a devastating impact on the Internet.”

Google also unveiled “information agents,” Gemini Spark, and background task tools. Pichai said AI Overviews now reach 2.5 billion monthly users; Gemini has 900 million. Similarweb says Gemini holds over a quarter of generative AI traffic, up from 7% a year ago. Alphabet expects $180 billion to $190 billion in 2026 capital spending, focused on AI chips and infrastructure. Last quarter, Google’s ad revenue rose nearly 16% to $77 billion.

Posted on 22 May 2026