Google AI Overviews answer the query above the classic blue links. Informational searches that used to send the click to your blog now end on the results page. Impressions in Search Console stay steady. Clicks and average position get worse on the queries the overview covers. That gap is the zero-click problem, and publishing more of the same thin posts makes it larger.
You cannot turn the overview off. You can become one of the few sources it quotes, and you can aim the queries you still own at pages that are not fully answered in two sentences.
Which queries you just lost
Split the query list into two piles.
| Query type | What the overview does | What you should publish |
|---|---|---|
| "What is X" | Answers it in full | A definition page with one extra depth the overview cannot fit: a table, a worked example, or a tool |
| "X vs Y" | Compares the obvious points | Your own comparison with prices, limits, and a recommendation |
| "How to X" with a short answer | Lists the steps | A step page that includes the failure cases and a screenshot |
| "X near me" or a product name | Often skips the overview | Keep these. They still send the click |
| A query only your product solves | Rarely summarized well | Build the tool page and point informational posts at it |
The last row is the one most sites skip. They write another definition and hope. The click comes back when the article hands the reader a job the overview cannot finish.
Direct answer: Treat AI Overviews as a citation fight on head terms and a click fight on job-to-be-done terms. Win the quote with an answer-first paragraph. Win the visit with a next step the summary does not replace.
How a cited paragraph is built
Look at a query you lost in Search Console: impressions up, clicks down, over the last 28 days. Open the page that used to rank. Rewrite only the top of it:
- First paragraph: the answer, in plain sentences, with the query's words in it.
- Second paragraph: the limit of that answer. What the overview leaves out.
- A table or a numbered list the model can compress but not replace.
- A link to the tool or the service that finishes the job.
- A date, so the passage looks maintained.
Then link sideways. An overview citation is more likely when your domain has two pages that agree with each other. This piece links to the INP article and to why Perplexity skips unquotable sites. Both support the same point: extraction, not volume.
The page that still gets the click
For Anavya, the query "what is generative engine optimization" can be answered in an overview. The query "why is my site not cited by ChatGPT" cannot be finished there, because the answer depends on that site's robots file, its first paragraph, and its INP. That is why the article ends at the scanner instead of at another definition.
Use the same pattern on your site:
- Informational post states the rule.
- The next link opens a page that runs the rule on the reader's URL, price, or template.
- About is for the person who wants to know who is doing the audit.
- Contact is for the site that needs the checklist applied by someone else.
What to watch for 30 days
Track three numbers, not one:
- Clicks on the queries that trigger an overview. Expect these to stay down. Do not panic-rewrite them every week.
- Citations or referral visits from chat tools, if your analytics splits them out.
- Clicks on the job queries that point at the tool. These should rise if the internal links are real.
If clicks on the tool page do not move, the article is still a dead end. Add the link higher. A citation that never offers the next step brings brand mentions and no pipeline.
The free scan is the next step on this site. It checks the same extraction and speed issues these three articles describe, on the URL you actually own.
