Do AI Overviews Cut Clicks on Arabic Health Searches? What Search Console Shows on Two Sites

On two Arabic health sites, the pages Google cited most did not lose clicks. Between 17 July and 16 October 2026, 22.5% of a large Arabic medical lab blog’s Google impressions came with a link inside an AI Overview or AI Mode answer. At the same positions, its most-cited third of pages earned 127.6 clicks for every 100 expected. The least-cited third earned 76.9.

A mid-size Gulf gastroenterology clinic pointed the same way, with wider error margins. Search Console only logs AI impressions for pages linked inside the answer, so the comparison is between pages cited often and pages cited rarely. That is a different test from the studies abroad, and it does not prove that AI answers add clicks.

What did we measure, and on which two sites?

We joined two Search Console reports, page by page, for two Arabic health properties over the same 92 days, 17 July to 16 October 2026. The Generative AI report gave each page’s AI-feature impressions. The Performance report gave its clicks, web impressions and average position. Dividing the first by the second gives each page’s AI share.

The two sites differ in scale. The first is a large Arabic medical lab blog run by a Saudi lab group and read from Saudi Arabia and Egypt to the Maghreb. It logged 41 million AI-feature impressions in three months, 22.5% of its web impressions. The second is a mid-size Gulf gastroenterology clinic, at 24.9%. Both clients approved our aggregated, anonymised use of their data.

AI share counts citations, not AI answers. Google logs an AI impression only when a page’s link appears inside the answer. A page can sit under an AI Overview all day and log nothing if the answer links someone else.

Google’s help page for the Generative AI report says the report draws on the Web search type of the main Performance report. AI impressions are part of your web impressions, so never add the two. We covered the report itself in our first reading of Search Console’s AI report on Arabic sites.

Before pulling the data, we wrote down what would count as a click penalty. It needed a most-cited to least-cited CTR ratio below 0.90, with its 90% interval excluding 1, in at least two position bands on both sites. Neither site met it, and both pointed the other way.

Do the pages Google cites most earn fewer clicks?

On both Arabic health sites, the pages Google cited most earned more clicks for their positions than the pages it cited least. On the lab blog, the most-cited third of 903 pages earned 127.6 clicks per 100 expected, against 76.9 for the least-cited third. On the clinic’s 103 pages, the gap was smaller: 104.7 against 81.6.

We treat the click index as the fairest single number here, because it compares each page with pages at the same average position on the same site. An index of 100 means a group earned exactly the CTR typical for its positions. The lab’s most-cited pages beat that by about 28%, its middle third sat at 93.0, and its least-cited pages fell about 23% short.

On the clinic, the middle third (100.6) and the most-cited third (104.7) sat close together, and the least-cited third did the falling. Its Arabic articles show the split more sharply: 70.6 for the least-cited third against 105.1 for the most-cited, with the middle at 105.2. With 71 article pages, a single strong page can move a third by several points.

At the same position, the most-cited pages still earned more

The lab blog’s most-cited pages kept their click advantage at similar average positions. At positions 4 to 6, they turned 1.51% of impressions into clicks, against 0.63% for the least-cited pages, a ratio of 2.39 (90% interval 2.02 to 2.82). At positions 7 to 10 the ratio was 1.47 (1.17 to 1.82).

In clicks per 10,000 impressions, that is 151 compared with 63 at positions 4 to 6, and 72 compared with 49 at positions 7 to 10. The middle third landed in between both times, at 97 and 69. The bars rest on 57, 165 and 269 pages in the first band and 244, 136 and 31 in the second, and neither interval comes near 1.

Across all 903 pages, the rank correlation between AI share and CTR was 0.68 (Spearman). By the usual rule of thumb, that counts as strong. The 880 Arabic article pages alone gave the same ratios. The more often Google linked a lab page in AI answers, the larger the share of its impressions that turned into clicks.

The clinic is thinner. Only one band held enough pages: at positions 7 to 10, its most-cited pages reached 0.60% CTR against 0.43%, a ratio of 1.38 with an interval from 0.48 to 2.51. That interval crosses 1, so we read the clinic’s all-pages result as a direction only. Its Arabic articles in the same band gave 1.72 (1.07 to 3.06).

Did CTR fall as AI exposure grew?

In our weekly data, CTR rose alongside AI share on both sites from late July to late September 2026. The lab blog’s AI share went from 19.9% in the week of 9 August to 26.0% by mid-September as its CTR climbed from 0.81% to 1.02%. In October both eased, to 24.3% and 0.88%.

The clinic moved further. Its AI share went from 18.4% in the week of 26 July to 33.8% in the week of 27 September, while its CTR rose from 0.55% to 0.82%. Average position improved on both sites over those weeks, from 6.64 to 6.13 on the lab blog and from 8.25 to 6.47 on the clinic.

Two things overlap this rise. Better positions lift CTR on their own, and part of the position gain can come from Google’s rule that links inside an AI Overview take the Overview’s position. The September 2026 spam update also ran from 24 September to 8 October, covered in our check of that update on Arabic sites. We cannot separate these effects in this data. Thirteen weekly points also describe one window, too short to call a trend.

Your Search Console already holds both reports. Show me which of my pages earn clicks when Google answers first

Global studies vs what we measured

The best-known studies outside the Arab world all find lower CTR when an AI Overview appears. Pew counted a result click on 8% of US visits with an AI summary against 15% without, and Ahrefs measured 58% lower position-1 CTR. None of them measures what Search Console’s AI report allows: pages cited often against pages cited rarely.

The studies agree on the direction and differ on the size, from 15.49% lower CTR in a second US agency’s study of 700,000 keywords to 58% at position one in Ahrefs’ data. None of them covers Arabic queries or an Arab market.

The closest match to our design is the 2026 update from the US search agency. On informational queries in 2025, its brands earned 2.07% CTR when cited inside the AI Overview and 0.94% when the Overview appeared without them. Its authors give the same warning we do: higher-authority brands are also more likely to be cited.

Both results can hold at once, because a page with a low AI share is not a page untouched by AI answers. Part of its impressions can come from searches where an answer appeared and linked someone else, the lowest-click case in every study above. Search Console can show you when your link sat inside an AI answer. It cannot show you when an AI answer appeared without you.

The French data carries a warning for health publishers. Health queries in that panel triggered AI Overviews 22.2% of the time, the highest rate in the study, and health CTR was 20.4% lower in the nine days after launch than in the 28 days before.

What this data cannot prove

Our data links being cited often with earning more clicks; it does not show that citations cause the clicks. It cannot say what any page would have earned with no AI answer at all, because neither site has a group of pages never shown in AI features. Four limits shape how far the result travels.

• We have no control group on either site. Every clinic page with at least 300 impressions had some AI impressions. The lab set also leans to winners: it holds only pages that made both 1,000-row exports, and the Performance export is the blog’s top 1,000 pages by clicks, so low-click pages are under-represented.
• Strong pages can win on both counts. A page Google trusts as a source is often the page searchers prefer to click. On the lab blog, 269 of the 301 most-cited pages sit at average positions 4 to 6, while 244 of the 301 least-cited pages sit at 7 to 10.
• Citations bend average position. Google gives every link in an AI Overview the Overview’s own position, and average position takes each page’s topmost position per query. Cited pages look higher than their blue links. That should drag their in-band CTR down, so this rule works against our result.
• Two sites and one window are a small base. The AI report has no query dimension and no split between AI Overviews and AI Mode, and 92 days on two properties do not cover all Arabic health search.

We are confident about the direction on the lab blog: two position bands, intervals well clear of 1 and a rank correlation of 0.68 across 903 pages. The clinic suggests the same direction, but its all-pages interval runs from 0.48 to 2.51, so treat it as a supporting reading.

What should Arab health publishers measure?

Track three things on your own site: AI share per page from the two Search Console reports, a position-adjusted click index per page, and a weekly site trend of AI share, CTR and average position. Together they show which pages Google uses as sources, whether those pages earn clicks for their positions, and what moved when traffic changes.

Read your average position with care on pages Google cites often. A cited page takes the AI Overview’s position, so it can climb in the report without moving in the blue links. The lab blog’s September gain in position arrived together with its rise in AI share.

Refresh all three every month and keep your exports, because your comparison gets stronger with every month of AI data you add. Search Console never sees visits from ChatGPT, the Gemini app or Perplexity, so pair it with tracking AI traffic in GA4.

How to run the comparison in Search Console

You need two exports and one spreadsheet, and if your property has the Generative AI report you can run the comparison in an afternoon. Export the Pages tab of both reports for the same dates and search type, join them by URL, split pages into thirds by AI share, and compare CTR inside position bands.

1. In your property, open Performance, then Search results. Set at least three months, keep the search type on Web, open the Pages tab and export it.
2. Open Performance, then Generative AI, with the same dates. Set the same search type you use in the Performance report (we used Web in both), open Pages and export. If the entry is missing, the property usually has too few AI impressions, or it is excluded from AI features.
3. Join the two files on page URL with XLOOKUP and drop URLs with a “#” fragment. Both tables stop at 1,000 rows, so on a large site you analyse the overlap. Values shown as “~” or “-” export as zeros; treat a zero position as missing.
4. Keep pages with at least 300 web impressions, so a single click cannot swing a page’s CTR.
5. Add AI share (AI impressions ÷ web impressions) and split your pages into three equal groups by it.
6. Band pages by average position: 1 to 3, 4 to 6, 7 to 10 and 11 to 20. In each band, sum clicks and impressions per group and divide, never averaging page CTRs. Skip cells with fewer than 10 pages.
7. For the click index, multiply each page’s impressions by its band’s overall CTR to get expected clicks, then divide actual by expected for each group.
8. Export the Dates tab of both reports every month and chart your weekly AI share against CTR and position.

What to do with pages Google rarely cites

For pages Google rarely cites, first check whether an AI answer appears for their main questions. If one does and links other sites, your page sits in the lowest-click position: seen in the results, skipped in the answer. Then fix the opening answer, give searchers a reason to click, and show medical credentials.

On the evidence so far, being left out of an AI answer costs more clicks than being in one. Search each page’s two or three main questions in Google and in AI Mode, in Arabic and in English, and note which sites the answer links. Our guide to how pages earn a place in AI Overviews covers the on-page side.

• Answer in the first lines: put the direct answer and the numbers readers came for (normal ranges, fasting hours, result times) in the first paragraph, in plain Arabic.
• Give searchers a reason to click. Test prices, booking, home sample collection and branch hours are things an AI answer cannot hand over. Transactional pages face this less: in February 2026 the US agency study found AI Overviews on 5% of transactional queries against 36% of informational ones.
• Show who checked the page. Google’s quality rater guidelines treat health as a “Your Money or Your Life” topic, so show a named author, a named medical reviewer with credentials, a review date and links to health authorities.
• Do not count on stars. Search Engine Roundtable reported on 7 October 2026 that Google had stopped showing review stars for healthcare pages, so your title and description now carry the click alone.

Do not prune a page for a low AI share alone, since a page that earns its expected clicks without citations is doing its job. A step-by-step healthcare SEO plan puts these fixes in order for clinics and labs.

Our read

Stop treating every AI citation on Arabic health content as a lost click. On our two sites, the pages Google cited most earned the most clicks for their positions, and the least-cited pages fell short. The pages to worry about are the ones that rank but rarely appear in the answer.

That changes what goes in your monthly report. Put AI share and the click index per page next to rankings, and judge a page by clicks per impression at its position, never by average position alone.

We would still not treat AI share as a target on its own. A citation earns its keep only when the page behind it offers something the answer cannot: a price, a booking, a result explained by a named doctor.

For Arabic health publishers, the first job is the least-cited third of pages that already rank in the top 10. Those pages keep the impressions and miss the clicks their positions should bring, so the rewrites start there.

FAQ

Do AI Overviews reduce clicks on health websites?

Studies abroad find lower CTR when AI Overviews appear: in France, health had the highest AI Overview rate (22.2%) and 20.4% lower CTR in the nine days after launch. On our two Arabic health sites, pages cited most often earned more clicks for their positions than pages cited rarely, 127.6 against 76.9 per 100 expected on the lab blog.

Why are my impressions up but clicks down in Search Console?

AI answers add impressions. Search Console logs one whenever your page’s link appears inside an AI Overview or AI Mode answer, even when nobody taps it. In March 2025, Pew found that only 1% of US visits with an AI summary included a click on a link inside it. Check your pages’ AI share before blaming rankings or content.

Can I see clicks from AI Overviews in Search Console?

No. The Generative AI report shows impressions only, for AI Overviews and AI Mode combined, by page, country, device and date. Clicks on AI links sit inside the normal Performance report’s Web search type, with no filter to separate them. The closest substitute is comparing pages by AI share, as this article does.

My average position improved but traffic dropped. Why?

Google gives every link in an AI Overview the Overview’s own position, which usually sits at the top of the results. A page cited in the answer can show a better average position without earning the clicks a real top blue link brings. Read your position next to AI share and your page’s position-adjusted click index.

Is it worth being cited in AI Overviews if people don’t click?

On our data, yes. The lab blog’s most-cited third earned 127.6 clicks per 100 expected for their positions, the least-cited third 76.9. A US agency study of 53 brands found 2.07% CTR in 2025 when cited and 0.94% when an AI Overview left the brand out. Google also says clicks from results pages with AI Overviews are higher quality.

The better question for Arabic health sites

Asking whether AI Overviews cut clicks treats every page the same. Ask instead which of your pages still earn the click when Google answers first, and which rank without ever being cited. Search Console now answers that page by page, and on our two sites the answer favoured the cited pages.

If you already have the Generative AI report, run the eight steps on the last three months and list the least-cited pages that rank in the top 10. That list is your rewrite queue. If the report is missing or nearly empty, start earlier: indexable pages that answer one question each, named medical review and fast mobile pages.

We can run the same join on your property. Run this Search Console comparison on my site