AI summaries appear in about 48% of search results, is the natural click-through rate enough?
The natural click-through rate is not enough when only looking at it. The public statistics put the reason for this matter very clearly: AI summaries have appeared in about 48% of search queries, and when they appear, the natural click-through rate decreases significantly compared to queries without summaries, and the page provides a comparison standard of 0.61% for queries with summaries and 1.62% for queries without summaries. That is to say, the decline in click-through rate is mixed with two things - your ranking changes, and the change in the form of the search results page. A single indicator cannot separate them.
What do the two numbers measure?
The rate of occurrence refers to the page form: how many queries will be occupied by AI summary at the top of the screen. The click rate refers to user behavior: how many people are willing to click in under the premise that you have already gained exposure. The former is an external variable, and the latter is the part that your content and structure can affect.
Put these two numbers together and you will get three actual conclusions:
- In information queries (‘What is X’ ‘How to choose X’), the summary can directly give the answer, and the motivation to click is naturally weakened;
- In transactional and operational queries (need parameter table, need template, need download, need to personally compare configuration), users still need to click in;
- When different intentions mix under the same keyword, the average hides the real change, and it is necessary to view by intention level.
Why click-through rate is no longer a single conclusion
| Observation scope | What can it explain? | What can mislead? |
|---|---|---|
| Natural click-through rate | The synthetic result of the change in the form of the result page and the attractiveness of the content | When the rate of abstract appearance increases, even if the ranking does not change, the click-through rate will fall |
| Brand word search volume | Whether the user remembers your name, not just absorbing the answer | Brand words are truncated or misspelled and scattered |
| AI recommended session | Whether the summary and Q&A tool really sends people to the site | The tool does not report the source, and the tags are incomplete |
| Lead generation and conversion sources | Did business value arise after the site visit | Only looking at the number of sessions will treat invalid traffic as performance |
Among these four items, the search volume of brand words is the most easily overlooked leading indicator. It measures whether there is an impression left after being cited, while click-through rate only measures whether there is a click this time.
The three types of perspectives that webmasters should supplement.
The first category is the demand-side aperture: divide the query into buckets of information type, comparison type, operation type, and transaction type, and separately view exposure and clicks. After bucketing, you will find that the summary truly consumes the traffic segment of the information type, and if your content structure is all written in the information type, it is equivalent to placing the entire site on the weakened segment.
The second type is the citation口径：Regularly use question sentences to search for your own site through Q&A tools and AI search, record whether it is cited, which paragraph is cited, and whether the site name is brought out. This type of observation cannot rely on the search console, and manual or sample checks with tools are needed, and the samples need to be fixed in order to compare.
The third type is the access口径：The behavior after arriving at the site. Whether the landing page answers the question mentioned in the summary, whether the lead entry is visible on the first screen, and whether there is a next step to continue reading in the content. The summary cannot replace clicks, but it will increase the threshold for clicks.
What settings can be配合 on-site
The site side is not just about 'installing a file and that's it', but also about aligning the three things of being accessible, citable, and承接.
AnQiCMS has several native settings that can be配合 on this chain. llms.txt will be automatically generated, allowing the large language model to directly read the site structure and key page entry during retrieval, which belongs to a 'referable' link; the sitemap (sitemap) is automatically generated and supports link push and active push to search engines, responsible for 'crawling'; the keyword library manages SEO keywords centrally, and anchor text management is responsible for stabilizing keywords to the corresponding page, these two determine whether they can handle after being referenced.
It should be reminded that these settings only reduce the threshold to a reasonable position without changing the form of the results page. The focus of observation should be on brand words and conversion sources, which is more valuable for decision-making than watching the fluctuations of click-through rate every time.
Common questions
Should the content be made shorter for easy abstract citation?
Don't shorten the main text just to be cited. A more effective approach is to make the first paragraph of each article able to stand alone as an answer, put the conclusion in the forefront, while retaining the details that can be clicked to verify, so that the abstract citation is your statement, and the user still needs to come and look at the complete basis.
How to count the search volume of brand words?
In the search console, filter the parts containing brand names by query terms, and then compare them with the site's own on-site search terms. When brand spellings have variants, they should be merged and counted together, otherwise they will be separated.
How often is the cited status sampled?
Sample by the core issues list, with the same group of questions fixed in each round, to avoid the results being uncomparable due to changing questions each time. The sample size does not have to be large, but the questions must be stable.