A business owner told me recently, “There’s no point doing keyword research anymore. Google’s AI answers everything, so nobody clicks.” I understand why people feel that way. The click numbers for some searches really have dropped hard.
But “nobody clicks” is not true. People still click when the AI answer can’t finish the job for them. They click to use a tool, to book a plumber, to compare prices, to see the latest version of something, or to read about someone’s real experience. Those searches are where your keyword research time should go first.
In this guide, let me explain what people mean by AI-proof keywords, which kinds of searches still earn clicks, and a simple method for finding and scoring them using your own Search Console data. I will also show you how to keep this list fresh, because what is safe today may not be safe in six months.
What “AI-proof” really means for a keyword
Let me be clear first: no keyword is fully AI-proof. Google can add an AI Overview to almost any search, and ChatGPT can answer almost any question. So I prefer the phrase click-resilient. It means a search where people still need to visit a website, even when an AI answer is on the page.
Ahrefs used the term “AI proof keywords” in a July 23, 2025 study of its own Search Console data. They looked for keywords where the click-through rate (CTR, the share of people who see your result and click it) held steady or grew, even though an AI Overview appeared. Out of 1,393 keywords with improved CTR, 97 (about 7%) also had an AI Overview on the page. The author calls the study small, so I treat it as a useful clue, not a rule.
“If you optimize your content well, it is still possible to improve CTR in AI Overview SERPs”
— Louise Linehan, Ahrefs, What We Learned From Studying Our Own “AI Proof” Keywords
The three patterns they found were keywords for free tools, keywords where the site moved up the rankings, and keywords where the AI Overview sat below position one. I build on those, and add the patterns that the wider data supports.
Why does this matter so much? Because the click loss on the other searches is real. Ahrefs’ February 2026 update found that, as of December 2025, an AI Overview was linked to a 58% lower CTR for the top organic result. I explain the full picture in my post on zero-click searches.
Seven kinds of searches where people still click
Here is the list I use. The question behind each one is simple: can the AI answer finish the job, or does the person still need a website?
| Search type | Why the AI answer can’t finish the job | Example |
|---|---|---|
| Tool and calculator searches | The person needs to enter their own data and get a result | “mortgage repayment calculator”, “free website traffic checker” |
| Transactional searches | They want to buy, book or sign up, which happens on a site | “buy running shoes size 11 wide” |
| Local service searches | They need a nearby business, opening hours, a phone number | “emergency electrician near me” |
| Commercial comparison searches | They want to see prices, photos and details side by side | “best CRM for small law firms pricing” |
| Fresh or live searches | The answer changes daily: stock, scores, events, deals | “Black Friday laptop deals this week” |
| Experience searches | They want a real person’s opinion, photos or story | “is [product] worth it after a year” |
| Branded and navigational searches | They want a specific site or login page | “[brand] customer portal” |
By contrast, the searches most at risk are simple definitions, quick facts and “how does X work” questions. An AI answer usually satisfies those completely.
What the data says about which searches avoid AI Overviews
The best data I have found on this is Ahrefs’ November 10, 2025 study of 146 million SERPs from September 2025. Overall, about 20.5% of searches showed an AI Overview. But the rate changed a lot by search type:
| Search type | Share with an AI Overview (Sept 2025) | What it means for click resilience |
|---|---|---|
| Navigational | 0.9% | Very safe, but you only win these for your own brand |
| Transactional | 2.1% | Very safe, and close to revenue |
| Shopping category | 3.2% | Safe from AI Overviews, but crowded with shopping ads and listings |
| Commercial | 4.3% | Mostly safe, high value |
| Local queries | 7.9% | Mostly safe; the map pack is the real competition |
| Branded | 13.1% | Fairly safe |
| Informational | 21.4% | At risk |
| Questions | 57.9% | High risk for clicks, good for AI citations |
Query length matters too: single-word searches showed AI Overviews 9.5% of the time, while searches of seven or more words showed them 46.4% of the time. So long, question-style keywords, which we were all told to target for years, are now the ones most likely to be answered on the results page.
Even when people do see an AI answer, they rarely click its links. Pew Research looked at the browsing of 900 US adults in March 2025. As PPC Land reported on July 23, 2025, users clicked a traditional result in 8% of visits with an AI summary, compared with 15% without one, and clicked a link inside the summary only about 1% of the time. More detail is in my post on what triggers AI Overviews.
Finding click-resilient keywords in your own Search Console
Industry data is useful, but your own data is better. Here is how I would do it with just Google Search Console and a spreadsheet. It takes about an hour.
- Open the Performance report and select Search results. If you are new to it, my guide to the Search Console Performance report covers the basics.
- Compare two periods. Pick the last three months and compare with the same three months a year ago. Turn on Clicks, Impressions, CTR and Position.
- Open the Queries tab and export it to Google Sheets or Excel.
- Remove tiny numbers. Filter to queries with at least 100 impressions in both periods. Ahrefs used the same 100-impression cut-off.
- Add a CTR change column. Use:
(new CTR − old CTR) / old CTR. - Add a position change column so you can separate “CTR fell because we dropped in rankings” from “CTR fell even though we held our position.”
- Sort into three groups: CTR stable or up (resilient), CTR down with stable position (likely hit by AI or other SERP features), and CTR down with position down (a ranking problem, not an AI problem).
- Tag each query by type using the seven types above. A simple regex filter helps here, such as
^(how|what|why|when|who|is|can|does)to pull out question queries. - Check the live SERP for your top 30 resilient and top 30 hit queries, and note which features appear. The next section explains what to look for.
When you finish, you usually see a pattern specific to your site. A service business often finds its “cost” and “near me” queries held up, while its “what is” blog queries fell away.
Reading the SERP for click-resilience signals
When you look at the live results for a keyword, these are the clues I check:
- Is there an AI Overview, and where is it? Ahrefs found AI Overviews sit outside position one about 9% of the time. Their small study saw better CTR when the AI Overview was lower down.
- Are there ads, a shopping carousel or a map pack? These signal buying intent, which is good, but they also push organic results down.
- Are the top results tools, product pages or forums? Google is telling you people want to do something or hear from real people, not read a summary.
- Do results show dates? Recent dates in the snippets suggest freshness matters, which AI answers handle less well.
- How long is the AI answer? A short, complete answer (“The boiling point of water is 100°C”) means few clicks. A vague or “it depends” answer leaves room.
A simple scoring sheet for click retention
Once you have a list, you need a way to rank it. I use a simple score out of 10. It is not scientific, but it forces you to look at each keyword the same way.
| Factor | 0 points | 1 point | 2 points |
|---|---|---|---|
| Task can only be done on a site | No | Partly | Yes (tool, booking, purchase) |
| AI Overview present | Yes, at the top | Yes, lower down | No |
| Your own CTR trend (year on year) | Down more than 30% | Down 0–30% | Flat or up |
| Business value of a visit | Low | Medium | High (lead or sale) |
| Freshness or experience needed | No | Somewhat | Strongly |
Keywords scoring 7 to 10 go to the top of your plan. Keywords scoring 0 to 3 may still be worth targeting, but for a different goal, which I cover below.
A worked example (hypothetical numbers)
Imagine a roofing company with two keywords:
- “how long does a roof last”: 2,400 impressions a month. AI Overview at the top. CTR went from 4.0% to 1.5%. Low business value. Score: 0 + 0 + 0 + 0 + 1 = 1.
- “roof repair cost [city]”: 600 impressions a month. No AI Overview, map pack showing. CTR went from 6.0% to 6.5%. High value. Score: 2 + 2 + 2 + 2 + 1 = 9.
The first keyword has four times the impressions, but it now brings about 36 clicks a month (2,400 × 1.5%), down from about 96. The second brings about 39 clicks (600 × 6.5%), and those visitors are much closer to calling. In this example, the smaller keyword is clearly the better use of the next month’s effort.
Your usual keyword tools still help with volume and difficulty. My guide to keyword research tools covers which ones show SERP features, which makes this scoring much faster.
Page formats that keep earning the click
Finding the keyword is half the job. The page also has to give people a reason to visit that the AI answer can’t. These formats work well:
- Free tools and calculators. This was the clearest winner in Ahrefs’ own data. Even a simple cost calculator for your service can do it.
- Pricing pages with real numbers. Ranges, packages and what changes the price. An AI answer gives a vague range; your page gives the specifics.
- Local service pages with photos of real jobs, service areas and a clear phone number. Our local SEO services focus on exactly these pages.
- Comparison tables people want to scan and come back to.
- Original data and case write-ups that can’t be summarized without losing their value.
- Templates, checklists and downloads people need to save or use.
Also, write titles and meta descriptions that promise what the AI answer can’t: “Instant quote calculator,” “2026 prices for [city],” or “Photos from 40 real jobs.” My guide to improving organic CTR has more on this.
Balancing click-resilient keywords with AI visibility
I don’t want you to drop informational keywords entirely. They still have a job. They just have a different one now.
The keywords that trigger AI Overviews are the ones where your page can be cited as a source. A citation doesn’t bring many clicks, but it puts your brand in front of the person at the moment they are learning. Later, when they search for a service or product, they may remember you. Google’s own documentation says a page simply needs to be indexed, and meet Search’s normal technical requirements, to be eligible as a supporting link:
“There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”
— Google Search Central, AI features and your website
So here is how I would split the work for most small businesses:
- Around 70% of effort on click-resilient keywords that bring leads and sales.
- Around 30% on informational topics where the goal is to be cited and remembered, measured by AI mentions rather than clicks.
The right split depends on your business, so treat this as a starting point. If you want to understand how AI Overviews and AI Mode differ in what they show, read my comparison of AI Mode vs AI Overviews.
Re-checking your list as AI Overviews spread
A keyword that is safe today may not stay safe. AI Overviews grew fast in 2025. Ahrefs tracked them on 7.6% of US keywords on March 13, 2025, and 16.48% by May (Ahrefs, May 13, 2025). Featured snippets fell sharply in the same period, as I explain in what happened to featured snippets in the AI era.
Here is the routine I suggest:
- Every month: check CTR for your top 20 money keywords in Search Console. Flag any that drop more than 20% while position holds.
- Every quarter: repeat the full export and scoring. Move keywords between groups as needed.
- After a big Google update: re-check the live SERPs for your top 30 keywords. The March 2025 core update was when AI Overviews jumped.
- Keep a simple log of the date you first saw an AI Overview on each key search. Over time, this shows you the trend for your own niche.
For broader numbers you can compare against, my AI SEO statistics page is updated as new studies come out.
Questions people ask about AI-proof keywords
Are any keywords truly AI-proof?
No. Google can add an AI Overview to almost any search. Some search types, like transactional (2.1%) and navigational (0.9%), rarely show one, according to Ahrefs’ September 2025 data, so they are much safer.
Should I stop writing informational blog posts?
No, but change the goal. Use them to earn citations in AI answers and build brand awareness, and judge them by mentions and assisted conversions, not just clicks.
Do local keywords still get clicks?
Mostly yes. Only 7.9% of local queries showed an AI Overview in Ahrefs’ September 2025 data, though the map pack takes many of the clicks.
What is the fastest way to find my click-resilient keywords?
Compare CTR in Search Console year on year, filter to queries with 100+ impressions, and look for queries where CTR held or rose while position stayed the same.
How often should I re-check?
Check your top money keywords monthly and repeat the full analysis quarterly or after a major Google update.
Building your first click-resilient keyword list
If you do one thing this week, run the Search Console comparison and sort your queries into the three groups. You will likely find a short list of keywords that still bring real visitors, and a longer list that has quietly stopped working. That tells you where to put your next month of effort.
If you would like us to run this analysis for you and turn it into a content plan, our SEO team does this for businesses every week. Get a free quote and we will show you which of your keywords still have room to grow.
Further reading on AI-proof keywords
- What Happened to Featured Snippets in the AI Era
- Claude Code for SEO: Practical Use Cases for Non-Developers and Teams
- Do AI Answers Cite AI-Written Content?
Sources for this AI-proof keywords guide
- Ahrefs – What We Learned From Studying Our Own “AI Proof” Keywords (Jul 23, 2025)
- Ahrefs – What triggers AI Overviews, 146M SERPs (Nov 10, 2025)
- Ahrefs – AI Overviews reduce clicks, 2026 update (Feb 4, 2026)
- Ahrefs – AI Overview growth after the March 2025 core update (May 13, 2025)
- PPC Land – Pew study: AI search summaries reduce link clicks (Jul 23, 2025)
- Google Search Central – AI features and your website (updated Dec 10, 2025)
