AI Overviews can answer a question before anyone visits your site. That does not mean SEO is dead. It means publishers and small businesses need to stop confusing search impressions with readers, and start giving people a reason to click.
Lynka Team · October 28, 2026 · 13 min read · Marketing
Follow Lynka on Google
Get new articles in your preferred sources.
Imagine writing a genuinely useful article.
Google finds it.
Google indexes it.
Google shows it to people searching for the topic.
Google might even use its information to answer their question.
And almost nobody visits your website.
That is the increasingly uncomfortable world of AI search traffic.
On October 26, 2026, The Verge published a detailed look at the damage and dispute surrounding Google's AI-powered search features and the publishers that depend on referral traffic. Read the reporting.
The debate is bigger than one company complaining about fewer clicks.
It is a question about the old bargain of the web.
Publishers make useful content.
Search engines help people find it.
Some of those people visit the original website.
Now the search engine can increasingly give them the answer before they need to leave.
The awkward part?
Your SEO can technically work while your traffic doesn't.
Google's AI Overviews place generated summaries into certain search results. They can combine information from multiple sources and provide links to supporting pages.
AI Mode is a more conversational search experience where someone can ask follow-up questions rather than just review a traditional list of results.
Google says both are built on its existing search infrastructure, with techniques that retrieve and use relevant web content to ground answers. Google explains the technology and publisher guidance here.
That is good for people who want a quick answer.
It changes the economics for websites that created the original information.
You used to compete to be the page someone clicked.
Now you may also compete to be information someone reads without clicking at all.
There is credible evidence that the presence of an AI summary changes user behavior.
In a Pew Research Center analysis published in July 2025, researchers studied browsing behavior from US adults who agreed to share their activity in March 2025.
They reported:
| Observed Google search page | Share with a traditional search-result click |
|---|---|
| With an AI summary | 8% |
| Without an AI summary | 15% |
The study also found that users clicked links inside the AI summary on about 1% of visits to pages with such summaries.
These are the results of a particular observed sample and period. They are not a universal click-through rate for every keyword or every country in 2026.
But the direction is difficult to ignore.
When the answer is already there, some people stop searching.
Which is exactly what a good answer is supposed to help them do.
That's the paradox.
Google has pushed back on claims that AI Search is broadly destroying web traffic.
In its own August 2025 statement, Google said overall organic click volume from Search to websites was relatively stable year over year and argued that the quality of clicks was improving. Google's position is here.
That claim and Pew's study are not necessarily direct contradictions.
They measure different things.
One can observe lower clicking on pages that display summaries, while overall search behavior, query volume and traffic mix change in other ways.
The important point for an individual business is this:
A platform-wide average does not tell you what happened to your page.
Your article might gain traffic.
It might lose traffic.
It might receive more impressions and fewer clicks.
You need your own numbers.
This is particularly painful for new SaaS websites and small publishers.
You open Search Console.
Your impressions went up.
That feels great.
You check clicks.
Almost nothing.
What happened?
Several possibilities:
The trap is assuming every impression means Google almost sent you a customer.
It doesn't.
An impression means the page appeared in a search result under Google's reporting rules.
It is a visibility signal.
Not a visitor.
Not a lead.
Definitely not revenue.
No.
But it makes generic informational blogs a weaker strategy than they used to be.
Consider a page called:
What is a CRM?
If the article offers the same definition everyone else does, a search summary can satisfy much of the query without a visit.
Now consider:
Why Your Sales Team Has 300 Leads but Nobody Knows Who to Call
Or a page that contains:
Those provide reasons to go beyond the summary.
This does not guarantee traffic.
It creates something worth visiting.
That distinction matters more than article length.
SEO advice has an unfortunate habit of turning useful observations into arbitrary rules.
Detailed articles can perform well.
So someone decides every article must have 4,000 words.
Then you end up with:
What is a sales lead?
followed by a doctoral thesis, a history of selling, and a section about the origin of the word "lead."
The reader wanted a definition.
The writer wanted a word count.
Google's 2026 guidance emphasizes valuable, unique, non-commodity content, not writing the longest possible version of common information. Google's Search Central guidance makes this explicit.
So be detailed when the problem needs detail.
Be short when it doesn't.
The goal is to resolve the reader's actual question better than the search result already did.
This is the question I would ask before publishing any article in 2026.
Imagine Google displays the first two paragraphs of your content as a summary.
Would the user still gain something by opening the page?
If not, the article might be too easy to replace.
Ways to make the visit valuable include:
Not "CRM improves efficiency."
Show how a lead moves from initial contact to qualification, deal and next action.
Take a position and support it.
For example:
The best CRM feature for a five-person team might be the one that stops them forgetting follow-ups, not the AI scoring dashboard.
That is more useful than repeating a vendor checklist.
A calculator or generator may solve a task more directly than an article describing how to do it.
If two products serve different customers, say so.
A trustworthy comparison may recommend the competitor.
For launch dates, pricing, API specifications or survey figures, link to real sources.
Not decorative citation bubbles that disappear when someone copies Markdown.
After answering the question, show the reader the useful action they can take.
Not a random CTA dropped into the last paragraph.
If you already have an established site, start with evidence from your own queries.
Look for:
Queries where Google already shows your pages.
Then ask:
A new blog post is not always the solution.
Sometimes the right action is improving the existing article.
Sometimes it is extending a tool page.
Sometimes the right action is doing nothing because the keyword has little business relevance.
Suppose your site has:
/tools/sales-goal-calculator
Then you publish:
/blog/sales-goal-calculator
Then another:
/resources/how-to-calculate-sales-goals
All three try to own the same basic search intent.
Now the site has three mediocre answers competing with itself.
Better:
One primary URL owns the tool intent.
Supporting content links to it.
The content cluster makes sense to a human and to a crawler.
This is not a magic ranking trick.
It is avoiding unnecessary confusion.
You will see terms like:
AEO
GEO
AI search optimization
Some of the underlying practices are useful.
But do not buy a mysterious service because it promises to make your brand "AI visible" using secret metadata.
Google's official guide says its generative search features still rely on core Search ranking and quality systems and that foundational SEO remains relevant. Read the Google Search guidance.
In plain English:
There is no guaranteed prompt that makes Google's AI recommend your site.
Anyone selling that certainty deserves a little suspicion.
An AI answer can explain how to calculate a sales target.
A calculator lets you enter your own numbers.
An AI answer can explain WhatsApp QR codes.
A QR generator produces the actual code.
An AI answer can summarize a lead-generation approach.
A working product helps manage the prospects afterward.
Those are different kinds of value.
Informational content helps people understand.
Interactive tools help people do.
For a small software company, the combination can be powerful.
That is why a useful free tool page can deserve more effort than ten generic blog posts.
No, but choose the news carefully.
A release article can capture fresh search demand before the results become crowded.
It can also become stale in 48 hours.
The best timely articles do two jobs:
If an AI model launches, do not just repeat the benchmark table.
Explain the business economics, adoption constraints, and what a team should test.
That gives the article a longer shelf life than a launch summary.
Traffic still matters.
But it should connect to something.
For a SaaS website, useful measures include:
| Signal | What it tells you |
|---|---|
| Search impressions | Google is testing visibility for a query |
| Organic clicks | People reached the website |
| Click-through rate | Searchers chose the result at the positions shown |
| Engaged visits | Readers found enough value to continue |
| Tool use | Someone did something useful |
| Pricing/product visits | The content generated commercial interest |
| Signups and qualified enquiries | Interest became a business opportunity |
There is no point celebrating 20,000 low-intent impressions if none of the readers care about your product.
There is also no reason to panic over a low CTR without checking average position and query intent.
A page ranking far down the results normally gets fewer clicks.
AI answers are one variable, not an excuse for every SEO problem.
This is the uncomfortable but useful lesson from the publishing industry's struggle.
If all your discovery depends on a platform you do not control, that platform can change the terms.
You do not need to abandon Google.
You do need other ways for people to remember you.
For a business, that could include:
The objective is not to become an influencer.
It is to stop every new visitor being someone you have to reacquire from scratch.
A small CRM company should not try to become a general encyclopedia of AI.
The relevant questions are the ones connected to customers, sales, lead generation, invoicing, and running a small business.
An article about a model release makes sense when we can explain what it changes for actual work.
A sales-goal calculator makes sense because a person has a calculation to complete.
A guide about lead follow-up makes sense because people lose real opportunities that way.
And when a reader wants to move from theory to doing, Lynka Sales, Lead Generation and the site's tools should be easy to discover.
The blog is not the product.
It's one of the ways people discover why the product exists.
If I were running a small SaaS website today, I would divide content work into three parts.
Look at Search Console queries that already show impressions.
Repair weak titles, missing explanations, unclear examples and mismatched intent.
Do not create a duplicate article because an existing page needs work.
Choose topics with:
No random content calendar filler.
Practical tools.
Good documentation.
Current comparisons.
Interactive examples.
Strong product pages.
A reason to return.
That is how a content site becomes something more than a source for someone else's answer box.
Google's job is to answer the searcher's question.
Your job is to create something worth visiting.
Those goals used to align more neatly.
Now they sometimes compete.
That isn't fixed by writing more words, buying a new SEO acronym, or adding a little robot icon beside your article title.
It is fixed, as far as any business can fix it, by creating value that survives the summary.
Google can rank your article and still send you almost nobody.
So give the reader a reason to want more than the answer.
A Pew Research Center study of March 2025 browsing activity found fewer traditional search-result clicks on observed pages with AI summaries than on pages without them. Results vary by query, market, ranking and time period.
No. Google says its AI search features still build on core Search indexing and ranking systems. Good technical SEO, useful content and accurate information remain important.
Possible reasons include average ranking position, search intent, title appeal, competing results, AI summaries and other features of the search results page. Inspect individual queries and pages before drawing conclusions.
Original insight, practical tools, current evidence, interactive material, comparisons and task-completion pages can give people reasons to visit beyond reading a short summary. None is a guarantee.
No. Group related queries by search intent and give one strong URL ownership of each topic. Improve an existing page when it already serves the intent.
No. Structured data can help explain a page, but it does not guarantee search features or citations in AI answers.
No. Preferred Sources is a user preference feature, not a general-purpose guaranteed ranking boost for every searcher.
Keep the next step connected
Bring prospecting, follow-up, sales and the work after the sale into one connected workspace.
Keep reading
Anthropic's new Claude Duet promises to work in the background, watch connected apps and bring important work to you. It also raises a more awkward question: how much supervision does an always-on assistant need?
AI agents are moving into real workflows fast, but workers are discovering a new job: supervising the agents. The important metric is not how much work an agent produces. It is how much useful work remains after monitoring, correction and cleanup.
Reddit stopped supporting onsite lead generation forms and paused ads that depended on them. Advertisers now have to send prospects to their own forms and measure conversions with Reddit Pixel or Conversions API. That changes more than the form.