If you search for ”how to fix a leaky faucet,” Google gives you a whole page of English tutorials on “faucet repair.” You’ve never typed the word “faucet” before, but the results are spot on. That’s what semantic search does—it no longer focuses on your keywords; it focuses on what you’re trying to say.
When Google launched BERT in 2020, the mechanism was still hidden behind the scenes, and most SEO professionals continued to stuff keywords as they had been doing. But now that ChatGPT has over 100 million monthly active users and AI Overviews are directly siphoning off clicks from 58% right in the search results (data from Ahrefs’ analysis of 300,000 keywords), the days of keyword stuffing are over.
But many people are still using 2018 strategies to approach SEO in 2026. This article aims to explain how search engines and AI currently ”understand” your content—and what you can do about it.
Search engines have long since moved beyond simply matching keywords.
The core of semantic search is simple: search engines convert web page content into mathematical vectors and then compare the distances between these vectors in semantic space. The phrases ”how to fix a leaky faucet” and ”repair dripping faucet” share almost no common words, yet their vector similarity can reach 0.89. Search engines can therefore determine that they refer to the same thing.
What does this mean? You no longer need to write three separate articles for ”cheap tickets,” ”low-cost flights,” and ”budget flights.” A single in-depth article covering the topic can address all of these terms. Conversely, if you’re still relying on keyword density for optimization, your page won’t even make it past the initial screening.
In its trend report released in June 2026, Ahrefs cited a statistic: traffic to informational blogs has fallen to 28% of its peak, while traffic to pure affiliate marketing sites has dropped by 66%. The reason is straightforward: when a user’s question can be answered directly by AI, Google has no reason to direct clicks to a purely informational website.
However, traffic isn’t declining uniformly. Websites that search engines ”recognize” as authoritative entities in a particular field are still seeing traffic growth. This is why entity optimization has become so important.
What is an entity? Why does it determine your AI visibility?
An entity is a specific thing recognized by a search engine: a person, a company, a product, or a location. Google’s Knowledge Graph contains billions of entities and their relationships; the price of the iPhone 17 Pro is $1099, its manufacturer is Apple, and Apple’s CEO is Tim Cook.
When your brand becomes a node in the knowledge graph, your content gains additional trust weight. More importantly, AI search engines (ChatGPT, Perplexity, AI Overviews) are entirely built on entity understanding. After analyzing 75,000 brands, Ahrefs found that the correlation coefficient between a brand’s mention frequency across the web and its AI visibility ranged from 0.66 to 0.71. For mentions on YouTube, the correlation was as high as 0.74.
This is far more predictive than traditional backlink metrics. In other words, the key factor determining whether you’ll be cited by AI isn’t the number of backlinks you have, but rather how many times the AI has encountered your brand in its training data—and in what context.
Let’s look at a specific example. Ahrefs itself has encountered this problem: There are a large number of old articles online that describe the company in different ways—some call it an SEO tool, others a data analytics platform, and still others competitive analysis software. For a large language model, these conflicting descriptions can be confusing: In what scenarios should it recommend Ahrefs?
Shifting from a Keyword-Centric Approach to a Topic-Coverage Approach
Since search engines rely on semantic understanding to match content, your strategy should shift from ”optimizing for individual keywords” to ”covering the entire topic space.”
How exactly do you do this? Let’s say you’re building a website related to cross-border payments. The old approach was to write a separate 800-word article for each of the following topics: ”Cross-Border Receipts,” ”International Payments,” and ”Overseas Receipts.” The new approach is to write a comprehensive guide that systematically covers the topic: comparisons of payment methods, fee analysis, processing times, compliance requirements, country-specific differences, and common pitfalls—a single piece of content that answers all the relevant questions users might search for.
The benefit of this approach is that your content can semantically match hundreds of different search queries, even if the exact terms used by users don’t appear a single time in your article. This is the core logic behind ”topic coverage takes precedence over keyword density.”
There’s a simple way to check if your coverage is sufficient: Open the top 10 competing articles and see which subtopics they cover in common, what questions they answer, and what data they use. Incorporate these elements into your own content. The underlying principle behind this approach is that semantic search ranking models evaluate the extent to which your content ”comprehensively” addresses users’ implicit needs.
Seven Things You Can Start Doing Right Away
First, standardize your brand’s business information.Your company name, product name, and description of core services must be consistent across your official website, social media, industry directories, and wiki-style platforms. Don’t describe yourself as a ”cross-border payment platform” in one place and an ”international settlement service provider” in another. For AI, the more inconsistent your descriptions are, the less certain it will be about which category to place you in.
Second, proactively fill in the gaps in your brand information.AI will piece together your brand profile from across the web. If the content on your official website isn’t specific enough, it will look for information on Reddit, forums, and blogs to fill in the gaps. Those sources may not be accurate. The solution is to write your own detailed FAQs, “How It Works” pages, and product comparison pages—filling the search results with specific descriptions and verifiable facts.
Third, use structured data to help machines understand you.Schema markup (structured data markup) is a tool that clearly communicates the semantic meaning of content to search engines. Use the Article schema for articles, the HowTo schema for tutorials, the FAQ schema for Q&A, and the Product schema for products. There is one point of contention: Some experiments have found that AI crawlers do not execute JavaScript, so if your schema is injected via JavaScript, the AI may not be able to see it. The safest approach is to output the schema directly in the HTML rendered on the server.
Fourth, create clusters of topics rather than isolated articles.A main pillar page links to multiple niche articles, creating a clear hierarchical structure. The anchor text for internal links should use descriptive language so that search engines can understand the thematic relationship between the linked page and the current page. This effectively tells search engines: “These topics are related, and I have systematic knowledge in this field.”
Fifth, strive to be cited.In ChatGPT’s sources, 43.8% refers to ”best-of” lists and buying guides. If you can produce comparative content that includes clear assessments and verifiable data, your content is much more likely to be cited by the AI. Be careful not to publish ”best-of” lists that rank your own site first on your own domain; Google is already cracking down on this type of self-promotion.
Sixth, expand your presence across more platforms.Brand mentions are the strongest predictive indicator of AI visibility. This means you need to do more than just SEO—you also need to focus on brand building, community management, content marketing, PR, and YouTube videos. The more your brand appears on YouTube, Reddit, and industry media, the more often the AI will encounter it in its training data, and the higher the probability that it will recommend your content.
Seventh, structure the content into ”atomic” units.Provide the answer at the beginning of each section, followed by context and an explanation. This is because AI systems, like human readers, focus their attention on the first few lines. A well-structured block of content that presents the answer upfront is easier to extract and cite than a paragraph that beats around the bush.
Don't Treat Schema Markup as a Panacea
Structured data is useful, but don’t rely on it too heavily. There are some misconceptions in the SEO community that need to be clarified.
Some people think that simply adding the FAQ schema will result in rich snippets and an immediate surge in traffic. The reality is that Google significantly tightened the criteria for triggering FAQ rich snippets in 2023, causing the FAQ rich results for a large number of websites to disappear entirely. Schema is a tool to help search engines understand content; it is not a ranking factor in and of itself.
Some people are fixated on the llms.txt file. After analyzing 137,000 websites, Ahrefs found that 2,8% of them had published an llms.txt file, but 97% of those had never been read by any AI tool within a month. Most of the requests to read this file come from link-preview crawlers like Slackbot. Google has also explicitly stated that this file is not required for a website to appear in its AI features. Instead of spending time on this file, it’s better to focus on improving content quality.
The Underlying Logic of Semantic SEO: Giving Machines No Reason to Skip Over You
Let’s get back to the most fundamental question: What does semantic search mean for your website?
This means that the search matching logic has shifted from ”word frequency” to ”meaning.” The logic for evaluating content quality has shifted from ”keyword coverage” to ”topic completeness.” The way trust is established has shifted from ”number of backlinks” to ”consistency of the brand entity across the entire web.”
The good news is that this system naturally rewards truly useful content. Under a semantic search framework, an article that is thoroughly researched, clearly structured, and answers real questions is more likely to be matched than a low-quality article stuffed with 30 instances of the same keyword.
The bad news is that if you’ve been sticking to the old methods, your traffic may already be declining. Moreover, this trend isn’t going to reverse—the share of searches covered by AI Overviews continues to expand, ChatGPT’s user base keeps growing, and traffic to traditional informational content will continue to erode.
The adjustments you can make aren’t actually that complicated: Stop writing duplicate content for keyword variations, and start providing in-depth coverage of topics. Stop focusing solely on backlink metrics, and start paying attention to your brand’s presence across the entire web. Ultimately, what search engines and AI are doing is trying to understand knowledge and relationships in the real world. The more clearly you can tell them ”who you are, what you do, and what you’re good at,” the more reason they’ll have to recommend you to people searching.
If you'd like to learn more about how AI Overviews are affecting search traffic, check out our previous analysis:AI Overviews Are Eating Into Your Search Traffic. For SEO strategies related to AI-generated content, you can also refer toThis article. For the basics of keyword strategy, we recommend checking outSteps for Implementing Google SEO Keyword TechniquesThe
English 
Chinese