--- title: "How to Get Cited in Google AI Overviews: 8 Weeks of Our Own Data" description: "We tracked 18 B2B buying queries weekly for eight weeks. An AI Overview appeared every time, and 54.7% of its citations also ranked in the organic top 10." author: "Ognjen Martinovic" datePublished: 2026-09-14 dateModified: 2026-09-14 readTime: "9 min" canonical: https://www.flowhivedigital.com/insights/get-cited-google-ai-overviews-saas --- # How to Get Cited in Google AI Overviews: 8 Weeks of Our Own Data **Key takeaways** - Across eight SERP captures between 19 July and 14 September 2026 we recorded 129 AI Overviews on 18 buying-intent B2B Webflow queries — an AI Overview appeared on every single capture, without exception. - Each answer names about seven distinct domains. 54.7% of those citations went to a domain that also ranked in that query's organic top 10 the same day, so roughly 45% did not — ranking is a strong signal, not a requirement. - Our own site earned zero citations on any query except our company name, despite 14 structured articles, valid schema on all of them and a machine-readable feed. Structure without rankings bought us nothing. To get cited in a Google AI Overview you have to be a source the answer can reach: a domain that either already ranks for the query or carries platform-level authority of its own. That is not a theory — it is what eight weeks of our own weekly measurements show. Since 19 July 2026 we have captured the AI Overview and its cited sources for 18 buying-intent queries in our niche, every week, through a SERP API. This article publishes what we found, including the part that reflects badly on us. We are a Webflow agency, not an SEO tool vendor. The dataset is small and narrow by design: 18 queries a B2B SaaS buyer actually types before hiring someone to build a website. Treat it as one niche measured carefully, not as the whole web. ## What we measured, and how Every Monday morning an automated cycle runs one live SERP call per tracked keyword through DataForSEO, United States, English, depth 30, with the AI Overview loaded in the same request. For each keyword we store three things: our own organic position if we have one, the full organic top 10, and the AI Overview's list of cited source domains. Eight of those captures exist so far, dated 19 July, 3, 10, 17, 24 and 31 August, and 7 and 14 September 2026. The keyword set grew from 13 to 18 over that period as we added queries, which is why the early captures carry fewer answers. One measurement note that matters for reading every number below. Google's AI Overview lists sources as domains, and the same domain is sometimes named twice in one answer. We count **distinct domains per answer**, so an agency named twice in one AI Overview counts once. On that basis the eight captures contain 129 AI answers and 950 domain-citations drawn from 169 different domains. This also makes our headline comparison deliberately conservative in the wrong direction — see the ranking section below. ## Every query we track has an AI Overview. All 129 of them. This is the finding we did not expect to be so absolute. Across 129 captured answers spanning eight weeks, we have never once recorded one of these queries without an AI Overview. Not on "webflow agency pricing", not on "wordpress to webflow migration", not on "questions to ask a webflow agency". Plenty of industry writing still frames AI Overviews as something that shows up on informational queries and leaves commercial ones alone. In our slice of B2B services that distinction does not exist. Every query on our list is a query somebody types while choosing a vendor, and every one of them returns a machine-written answer above the results with roughly seven companies named in it. The practical consequence is simple. If you sell B2B services and you are planning your content around the ten blue links, you are planning around the second thing on the page. ## Who actually gets cited 169 domains earned at least one citation. Ten of them took 35.6% of everything, and 51 domains appeared exactly once. It is a concentrated head with a long, thin tail — which means the realistic goal is not "get cited sometimes" but "become one of the twenty or so domains this niche's answers keep returning to". Two things in the table are worth sitting with. The most-cited domain in our entire dataset is a competing Webflow agency, not a publisher and not a directory. And the second is YouTube, on queries about hiring an agency, where nobody buys anything from a video. Our own line is in the table too. All seven of our citations came from a single query — our own company name. Table 1: Most-cited domains across 129 AI Overviews, 19 July – 14 September 2026 (source: our own DataForSEO captures) | Domain | Answers citing it | Share | What it is | | --- | --- | --- | --- | | flowout.com | 57 | 6.0% | Webflow agency | | youtube.com | 51 | 5.4% | Video platform | | webflow.com | 44 | 4.6% | The vendor itself | | reddit.com | 34 | 3.6% | Community | | brixtemplates.com | 28 | 2.9% | Template shop | | skrol.agency | 27 | 2.8% | Webflow agency | | lscreativeagency.com | 25 | 2.6% | Webflow agency | | help.webflow.com | 25 | 2.6% | Vendor documentation | | joinamply.com | 24 | 2.5% | Webflow agency | | flowninja.com | 23 | 2.4% | Webflow agency | | flowhivedigital.com (that's us) | 7 | 0.7% | All seven on our brand name | ## How much does ranking still matter? For every answer we compared the domains the AI Overview cited against the domains ranking in that same query's organic top 10 on that same day. Across all eight captures, **520 of 950 citations — 54.7% — went to a domain that was also in the top 10**. The other 45% came from somewhere else: deeper in the results, or off the results page entirely. For context, Ahrefs analysed 863,000 keywords and 4 million AI Overview URLs and found that 38% of cited _pages_ also appeared in the top 10 for the same query, down from 76% in their July 2025 study (reported by Search Engine Journal, 2 March 2026 — via WebSearch). Our 54.7% is higher, but it is not a like-for-like number and we will not pretend it is: we match on domain, they match on page. A domain match is the easier test to pass, so our figure is an upper bound on the same idea. Read the two together as agreeing on the shape — ranking helps a lot and decides less than it used to — rather than disagreeing on the size. Within our own eight weeks the figure moved between 53% and 59% and then dropped to 47% on 14 September, the lowest we have recorded. One week is not a trend and we are not going to call it one. We will keep publishing the number either way. Table 2: Share of AI Overview citations that also ranked in the same query's organic top 10, by capture date | Capture date | AI answers | Domains cited | Also in top 10 | Share | | --- | --- | --- | --- | --- | | 19 Jul 2026 | 13 | 82 | 44 | 54% | | 3 Aug 2026 | 15 | 101 | 60 | 59% | | 10 Aug 2026 | 17 | 119 | 70 | 59% | | 17 Aug 2026 | 16 | 133 | 70 | 53% | | 24 Aug 2026 | 16 | 125 | 72 | 58% | | 31 Aug 2026 | 16 | 124 | 68 | 55% | | 7 Sep 2026 | 18 | 139 | 76 | 55% | | 14 Sep 2026 | 18 | 127 | 60 | 47% | ## What did not work for us Here is the part most articles on this subject leave out, because most of them are written by people selling the fix. Over the same eight weeks we published structured, long-form articles on the exact queries we track. Every one carries valid **BlogPosting** and **BreadcrumbList** markup — plus **FAQPage** wherever the article has a visible FAQ — verified on the live page after publish. The site serves an **llms.txt** index and a Markdown twin of every article at the same URL plus **.md**. GPTBot, ClaudeBot and PerplexityBot all receive a 200 — we re-checked on 14 September 2026, and our robots file blocks nothing. Result after eight weeks: **zero citations on any query except our own company name.** We are not concluding that schema and machine-readable feeds are worthless. We are concluding something narrower and better supported: on their own, on a site with no rankings, they did not produce a single citation for us in eight weeks. The variable we are missing is not structure. It is that we sit outside the top 30 on 17 of the 18 queries, so the answer has no reason to reach us. If someone tells you that adding FAQ schema will get you into AI Overviews, ask them for their before-and-after on a site that does not already rank. We ran that experiment on ourselves and the answer was no. ## Step zero: can the crawlers reach you at all? Before any of the above is worth discussing, check the boring thing. We run this on every site we take over and it fails more often than it should. - Request one of your own article URLs with the user agents **GPTBot**, **ClaudeBot** and **PerplexityBot**, and confirm each gets a 200, not a 403 and not a challenge page. - Open **/robots.txt** and read every Disallow line. Bot-management rules at the CDN can also block these agents without a single line in robots.txt, so test the agents, do not just read the file. - Confirm the page returns its content in the HTML, not only after JavaScript runs. None of this earns you a citation. All of it can silently prevent one. ## What we are doing next, and what we would tell you to do Our own read of this data, applied to our own roadmap: **Earn the ranking first, then the structure pays.** 54.7% of citations came from top-10 domains. That is the single largest lever in the dataset and it is the ordinary SEO job. The AEO work is what converts a ranking into a citation; it does not substitute for one. **Go where the answers already go.** YouTube and Reddit together account for 9% of every citation we have recorded, on queries about hiring an agency. Those are not our channels today. The data says they should be somebody's. **Write so a paragraph survives being lifted out.** An AI answer quotes a fragment, not a page. Every section here opens with a claim that stands up alone, with the number attached to it, because that is the unit that gets extracted. **Publish the number that makes you look bad.** There is exactly one thing in this niche that a competitor with more budget cannot copy: first-party measurement, published with the failures in it. That is the whole reason this article exists — and if it earns us a citation, that will itself be a data point we report. If you want the underlying reasoning on what a Webflow build should cost or how to vet the agency doing it, we have written both up: [what a Webflow agency actually costs](https://www.flowhivedigital.com/insights/webflow-agency-cost) and [12 questions to ask before you hire one](https://www.flowhivedigital.com/insights/questions-to-ask-webflow-agency). **We publish our own numbers. Ask us for yours.** AI-search visibility is part of every rebuild we do — crawler access, schema, a machine-readable feed and the weekly tracking that tells you whether any of it moved. [See our pricing](https://www.flowhivedigital.com/pricing)[Talk to us](https://www.flowhivedigital.com/contact) ## FAQ **How do you get cited in a Google AI Overview?** In our data the strongest observable precondition is reach: 54.7% of the citations we recorded went to a domain that also ranked in that query's organic top 10 the same day, and most of the rest went to platforms with authority of their own, such as YouTube, Reddit or the vendor's documentation. Structure your pages so a single paragraph answers a single question and can be lifted out intact, keep AI crawlers unblocked, and put the ranking work first. We have not found a shortcut that skips the ranking, and we tested one on ourselves for eight weeks. **Does ranking in the top 10 still matter for AI Overviews?** Yes, and less than it used to. Our own captures put 54.7% of citations on domains that also ranked top 10 that day. Ahrefs, working from 863,000 keywords and 4 million AI Overview URLs, put the page-level figure at 38%, down from 76% in July 2025 (Search Engine Journal, 2 March 2026). Different units, same direction: a top-10 position is the best predictor available and no longer a gate. **Does adding FAQ or Article schema get you cited?** Not on its own, in our experience. Every article we published in this period carries valid BlogPosting and BreadcrumbList markup, plus FAQPage where it has a visible FAQ, verified on the live page, plus an llms.txt index and a Markdown twin of every post. Over eight weeks that produced zero citations on any query except our own brand name. Schema is worth doing and it did not, by itself, move this metric for us. **How often do AI Overviews appear on B2B buying queries?** On the 18 B2B Webflow and SaaS-website queries we track, an AI Overview was present on all 129 answers we captured between 19 July and 14 September 2026 — a 100% rate in that set. This is a narrow, deliberately commercial keyword list in one niche, so do not read it as a web-wide figure; do assume that in a comparable services niche the AI answer is always there. **How do I check whether my site is blocking AI crawlers?** Request one of your own pages with each user agent — GPTBot, ClaudeBot, PerplexityBot — and confirm a 200 rather than a 403 or a bot-challenge page. Read every Disallow line in your robots file as well, but test the agents rather than trusting the file: CDN bot-management rules can block these crawlers without appearing in robots.txt at all.