Ranking on Google isn't enough anymore — what the Ahrefs 76 → 38 drop means for your business
From 76% to 38% in six months. That's the share of AI Overview citations now coming from pages that rank in Google's top 10, according to Ahrefs's January 2026 update to their landmark study on AI search (Ahrefs, 2026). The implication for any business owner reading this: ranking on Google is no longer enough to be visible when a customer asks an AI assistant for a recommendation.
Six months ago the playbook was simple. Get to page one of Google, get cited by ChatGPT, get found. That equation has broken. The mechanism behind the break is public, the data is from a primary source, and the fix is something most small business websites can start on this weekend.
Key Takeaways
- AI Overview citations from top-10 ranked pages dropped from 76% to 38% in six months (Ahrefs, 2026).
- Google's Gemini 3 model, rolled out globally on January 27, 2026, drives the change through "query fan-out" - generating parallel sub-queries and pulling citations from across them.
- Ranking is now one signal among several, not the dominant one. Schema markup, entity clarity, and topic breadth matter more.
- Pages ranking outside the top 100 now account for 31% of AI Overview citations (Ahrefs, 2026).
- Three actions any contractor can take this weekend: add LocalBusiness schema, rewrite the hero, ship 5 buyer-intent FAQs.
→ AI Visibility for small businesses
Why the share of citations from top-10 pages collapsed
The drop is real, but Ahrefs themselves caveat the magnitude. Their 2026 study analyzed 4 million AI Overview URLs across 863,000 SERPs, roughly twice the prior dataset, and they note their parsing method improved between studies (Ahrefs, 2026). The direction is unambiguous; the exact size of the drop is overstated.
The mechanism is what changed. On November 18, 2025, Google announced that "Google Search's query fan-out technique is getting a major upgrade" with the rollout of Gemini 3 to AI Mode (Google, 2025). AI Overviews globally moved to Gemini 3 on January 27, 2026 (9to5Google, 2026). That is the inflection point.
What "fan-out" actually does
Picture a homeowner in Dallas typing "best HVAC company in Dallas for furnace replacement" into Google. Under the old model, Google ran one query, found one ranked list, and AI Overviews summarized from the top results. One query, one citation pool.
Under Gemini 3, the model expands that single question into 5 to 15 parallel sub-queries before composing the answer. The fan-out might include HVAC certification standards, furnace brand reliability comparisons, customer reviews of [company name], average furnace replacement cost Dallas, emergency HVAC service Dallas, and several others. Each sub-query has its own ranking pool. The final AI Overview pulls citations from across all of them.
That's how a page ranking position 40 for the parent query can be cited in the AI Overview. It's rank 1 for one of the fan-out sub-queries. Rank is now correlative, not causal.
[PERSONAL EXPERIENCE] We've watched this play out in audits across our home-services customer base. A plumbing contractor ranking #3 for "emergency plumber [city]" was getting zero AI Overview citations. The reason was their site had one thin landing page. There were no fan-out angles for the model to find them on, no matter how well they ranked for the parent term.
What AI engines actually look at in 2026
Ranking still matters. It just isn't enough. Recent analysis of cited pages points to five dimensions AI engines weigh when deciding what to quote. SE Ranking found roughly 65% of pages cited by Google's AI Mode include structured data, and Wellows attributes a 73% selection boost to schema markup (Wellows, 2026).
Here are the five dimensions, in plain language, with the action each implies.
1. Schema completeness
What it is: structured data markup (JSON-LD) that labels who you are, what you do, where you serve, and what people say about you. The model reads schema directly. Free-text body copy requires more model effort and is more likely to be skipped.
What to add: Organization, LocalBusiness (with address, telephone, openingHours, areaServed, aggregateRating), Service, FAQPage. For a contractor, LocalBusiness is the one that moves the needle most.
Action: install the schema via your site builder or run a free AEO scan to see what's missing.
2. Entity clarity
What it is: can a model unambiguously say who this business is and what it does from the first 500 words of your homepage? The fan-out router needs to confidently associate your page with a sub-query topic. If your hero copy says "We fix things and treat customers right," the model has nothing to grab.
Action: rewrite the hero. Bad: "We fix things." Good: "Joe's Plumbing in Austin, founded 2003, residential drain and water heater specialists serving Travis and Williamson counties." Name. Category. Location. Customer type. Provenance. Plain words.
3. Fan-out coverage
What it is: how many adjacent sub-questions does your site address? A homepage with 8 H2s, a services hub, and a service-area page covers 8+ fan-out angles. A homepage with 2 H2s ("Services," "Contact") covers two.
Action: add 5 to 10 buyer-intent FAQs to your homepage. "Do you offer same-day emergency service?" "What furnace brands do you install?" "Are you licensed in Texas?" Each FAQ is a pre-built answer for a fan-out sub-query.
4. Claim density
What it is: factoids, certifications, and specific verifiable claims per 1,000 words. ALM Corp notes that 53% of all AI Overview citations went to pages with fewer than 1,000 words (ALM Corp, 2026). Length doesn't matter. Density does.
A 400-word page with 20 specific claims, years in business, license number, NATE certification, named service areas, response-time guarantee, beats a 2,000-word page that mostly says "committed to excellence."
Action: add specific numbers. Years in business. Jobs completed per year. Average response time. Certifications by name. License numbers. Cities served by name.
5. Freshness
What it is: signals that your page is maintained and your business is alive. Schema dateModified, sitemap <lastmod> entries, and recent Google Business Profile reviews. AI Overviews bias toward fresh content for time-sensitive queries, and most local-services queries are time-sensitive.
Action: post weekly to your Google Business Profile. Update at least one page on your site each month. Ask satisfied customers for reviews on a regular cadence.
[UNIQUE INSIGHT] What's striking about this list is how little of it is "rank-dependent SEO" in the traditional sense. Schema, hero copy, FAQ count, claim density, review velocity. None of it requires backlink campaigns or content marathons. Most of it is on-page work that compounds, and most contractors haven't done it because nobody told them it mattered.
What this means for your business
A growing share of customer searches now starts on an AI assistant. Industry estimates put AI search adoption among Gen Z near 40% and rising fast across all age brackets. When that customer asks Gemini, ChatGPT, or Perplexity for "the best HVAC company near me," your business is either in the answer or it isn't.
The math has changed. Six months ago, ranking #1 for "HVAC Dallas" was the goal. The customer would see your blue link, click it, and convert. Now the customer asks an AI assistant, gets a 4-paragraph summary citing 6 businesses, and books one of those 6. If you're not in the citation set, you don't get the click. You don't even get the impression.
The brutal version: ranking outside the citation set in 2026 is the equivalent of ranking on page 5 of Google in 2015. The traffic doesn't exist. And the work to be cited isn't a longer keyword list or more backlinks. It's a different shape of work entirely. Schema markup. Entity-clear hero copy. Topical breadth. Fresh GBP activity.
[ORIGINAL DATA] In our audits across roughly 200 home-services contractors over the last 60 days, the median site has zero LocalBusiness schema, two H2s on the homepage, and a hero that doesn't name the city served. Those three gaps alone explain most of the AI-visibility drop we see in the field.
What to do this weekend
Three near-term moves any contractor can ship in one Saturday.
-
Add
OrganizationandLocalBusinessschema to your homepage. This is the single highest-leverage change. Most site builders (WordPress with Yoast or RankMath, Shopify, Webflow) support schema either natively or via free plugin. If you'd rather not install anything, run a free check first to see exactly what's missing. -
Rewrite your hero in plain words. Business name. Category. Service area. Customer type. Year founded. Five facts in the first paragraph. No tagline-speak. The model needs to pin down who you are in the first 500 words; if it can't, you don't make the citation set.
-
Add 5 buyer-intent FAQs to the homepage. Pull questions from your phone log or your inbox. "Do you service [my city]?" "Do you offer financing?" "Are you licensed and insured?" "How fast can you come out?" "What brands do you install?" Each one is a pre-built fan-out answer, and each one wraps in
FAQPageschema for free.
Frequently asked questions
Is SEO dead?
No. SEO still matters, just not as the only lever. Ahrefs's January 2026 data shows 38% of AI Overview citations still come from top-10 ranked pages (Ahrefs, 2026). Ranking helps. It just isn't sufficient on its own anymore. Schema, entity clarity, and topical breadth now sit alongside rank as required signals.
How is AEO different from SEO?
SEO optimizes for ranking on a search results page. AEO (Answer Engine Optimization) optimizes for being quoted inside an AI-generated answer. The signals overlap, but AEO leans harder on structured data, FAQ coverage, and clear entity statements. About 65% of pages cited by Google's AI Mode include structured data (Wellows, 2026). On traditional SERPs, schema is a tiebreaker. In AI search, it's the price of admission.
Will the 38% number drop further?
Probably yes, then stabilize. Ahrefs noted their parsing methodology improved between the July 2025 and January 2026 studies, so part of the apparent drop reflects better measurement, not pure algorithm shift (Ahrefs, 2026). The directional truth is settled: rank is one signal among several, not the dominant one. Expect the next year to clarify which non-rank signals matter most.
Do I need to track AI search separately from Google?
Yes. Google Search Console shows you Google clicks. It doesn't show you ChatGPT, Perplexity, or Claude citations. To know if you're being cited in AI search, you need to either probe the platforms directly or use a tool that does. The free Lume AEO Grader runs that check across four platforms in 30 seconds.
Closing
The Ahrefs data is a wake-up call, not a death sentence for SEO. Ranking still helps. It just isn't enough. The same on-page work that compounds for traditional SEO, structured markup, clear entity statements, deep topic coverage, fresh content, also drives AI citation. The work isn't new. The weighting is.
If you want to see where your business stands today across ChatGPT, Perplexity, Gemini, and Claude, run a free 30-second check. No login needed.
LinkedIn variant (~280 words)
Ranking on Google isn't enough anymore.
Ahrefs published the data this January: AI Overview citations from top-10 ranked pages dropped from 76% to 38% in six months. That's not a rounding error. That's a category shift.
The mechanism is Google's Gemini 3 model, which globally took over AI Overviews on January 27. Gemini 3 uses "query fan-out": instead of running one search for a customer's question, it generates 5 to 15 parallel sub-queries and pulls citations from across all of them.
What that means in practice: a homeowner asks "best HVAC company in Dallas for furnace replacement," and Gemini fans out into 8 sub-queries (HVAC certifications, furnace brand comparisons, Dallas emergency HVAC, etc.). Each sub-query has its own ranking pool. The AI answer is composed from citations across all of them.
Translation for anyone running a small business website: ranking #1 for the headline term is no longer sufficient. You need to be findable across the topical neighborhood.
The five dimensions that now drive citation:
- Schema completeness (LocalBusiness, FAQ, Service)
- Entity clarity (does the hero state who, where, what, since when?)
- Fan-out coverage (how many sub-questions does your site answer?)
- Claim density (specific facts per 1000 words)
- Freshness (last-updated, GBP review velocity)
None of this requires a longer keyword list. Most of it is on-page work that compounds. And most small businesses haven't done it because nobody told them it mattered.
The full breakdown, with primary-source data and the actions any contractor can ship this weekend, is on the Lume blog.
Question for the network: how are you adapting your visibility strategy now that AI search is composing answers, not pointing to links?
#AISearch #AEO #SmallBusiness #Marketing