Do AI Overviews hurt my traffic?

For informational queries where users find their answer in the AIO, click-through rates have dropped 25-40%. However, sites cited inside AIO see higher brand visibility and higher CTR than position 1-3 organic results. Being cited beats ranking #1.

BLOG

AI SEO 2026: How to Rank in Google AI Overviews (AIO)

Q: What is Google AI Overview (AIO)?

AI Overview is the AI-generated summary Google displays at the top of search results for many queries. It synthesizes answers from multiple sources and cites 3-8 pages as references. AIO now appears for 60%+ of informational queries as of April 2026.

Q: How do I get cited in AI Overviews?

Structure content around direct question-answer pairs, demonstrate E-E-A-T with clear author bios and expertise signals, add new information that isn't already in the index (Information Gain), use FAQ and HowTo schema, and make sure individual paragraphs can stand alone as answers.

Q: What is Information Gain?

Information Gain is Google's measure of how much new information your page adds beyond what's already indexed. Pages that synthesize existing information score low. Pages with original research, first-hand data, primary sources, or unique perspectives score high. Information Gain has become a meaningful ranking factor in the AI era.

April 17, 2026 · 17 min read

A home improvement blog I follow used to get 180,000 monthly clicks from a single "how to fix a leaky faucet" article. In Q1 2026, that dropped to 48,000. The article still ranks #2 on Google. What changed is that Google now answers the question directly in an AI Overview at the top of the page, with a step-by-step fix synthesized from five different sources. The site gets cited as one of the five. The user doesn't click.

That's the 2026 reality for informational content. AI Overviews appear for more than 60% of informational queries, and they're eating click-through rates. But they're also creating a new winner: sites that get cited inside the AIO itself. A citation in an AI Overview shows up above any organic result, includes your brand name, and sends more qualified traffic than position 1 used to. The game changed, but it didn't end. Here's how to win the new one.

What AI Overviews Actually Do

When you query Google for something informational ("how to fix X", "what is Y", "why does Z happen"), the AI Overview engine does four things:

Decomposes your query into sub-questions (e.g., "how to fix a leaky faucet" becomes "what causes leaks," "what tools do I need," "step-by-step fix").
Retrieves relevant passages from its index using a hybrid of semantic search and traditional ranking signals.
Synthesizes an answer using a Gemini-based model, citing the sources inline.
Displays the answer at the top of the SERP, with source citations users can click to explore.

The retrieval step is where SEO gets interesting. The model doesn't just pull from the top 10 organic results. It pulls from whichever pages contain the best passage-level answers to each sub-question, which means pages ranking at positions 15-40 can get cited in an AIO if they nail a specific sub-question better than the top 10 do. Roughly 40% of AIO citations come from pages outside the top 10. That's the opening.

The CTR Reality: What Traffic Actually Looks Like Now

Here's what changed in click-through rates across informational queries since AI Overviews scaled:

Scenario	CTR Before AIO	CTR With AIO	Change
Position 1 (no AIO shown)	28%	28%	Unchanged
Position 1 (AIO shown above)	28%	11%	-61%
Cited in AIO (not ranking top 10)	1-2%	4-6%	+200%
Cited in AIO AND ranking top 3	28%	18%	-36%
Not cited, not top 10	0.5%	0.2%	-60%

The pattern: being cited in the AIO is the new "position zero" and it's worth more than organic rankings for informational queries. Commercial and transactional queries are less affected because AIO triggers less frequently for them, but the direction of travel is clear. Optimize for citation, not just rank.

The Five Signals That Drive AIO Citations

1. Passage-Level Answerability

AIO grabs individual paragraphs, not whole articles. If your "how to reset a router" article buries the actual answer in paragraph 12 after 11 paragraphs of preamble, AIO won't cite you. If paragraph 2 starts with a clear, standalone answer, it will.

The fix: structure content so each section's opening paragraph is a self-contained answer. Someone reading only that paragraph, without context, should get a complete answer to the section's question. Elaboration comes after. Here's the pattern:

H2: How to reset a Wi-Fi router

P1 (the answer): To reset a Wi-Fi router, locate the
recessed reset button on the back of the device, press
and hold it with a paperclip for 10 seconds until the
power light flashes, then release. The router will
restart and return to factory settings within 60 seconds.

P2+ (the context): The reset button is distinct from the
power button. It's usually recessed to prevent accidental
presses. Holding for less than 10 seconds triggers a
partial reboot that doesn't clear saved settings...

That P1 is the citation bait. AIO grabs it, attributes it to you, and users who want more detail click through to P2+.

2. Schema Markup (The API Layer for LLMs)

Schema markup became table stakes in 2026. It's how AI Overviews (and LLMs indexing the web generally) understand what your content is about without having to parse free-form prose. The schemas that matter for AIO citation:

FAQPage — explicitly tags question-answer pairs. Google's AI Overviews demonstrably prefer pages with FAQPage schema for "how" and "what" queries.
HowTo — tags step-by-step instructions. Ideal for procedural queries.
Article with author, datePublished, publisher — establishes E-E-A-T signals the AI weighs.
Product with reviews, price, availability — for commercial queries.
BreadcrumbList — helps the AI understand site structure and topical hierarchy.
Organization with sameAs links to social profiles — confirms entity identity.

If you're not shipping schema today, start there. A schema markup generator can draft valid JSON-LD blocks for any page type. Validate with Google's Rich Results Test before deploying.

3. E-E-A-T Signals (Experience, Expertise, Authority, Trust)

Google's AI Overviews lean heavily on author and publisher signals. An anonymous page on a domain with no author bios, no "About" page, and no visible expertise is radioactive for AIO citation. A page with a bylined author, author schema, linked LinkedIn profile, and a demonstrable track record in the topic is citation-ready.

Concrete actions:

Every article needs a real author byline with a linked author bio page.
Author bio pages need: credentials, relevant experience, links to social profiles, and ideally to the author's other work.
Add author schema on articles: {"@type":"Person","name":"...","url":"...","sameAs":["..."]}
Link out to primary sources and authoritative references. Pages that only cite themselves look like synthesized content.
Keep a visible "Last updated" date. Stale content gets cited less.

4. Information Gain

This is the big one. Information Gain is Google's signal for "how much new information does this page add that isn't already in the index?" A page that perfectly restates what three other pages already said has low Information Gain. A page with original research, first-hand data, unique angles, or primary-source analysis has high Information Gain.

In the AI era, synthesizing the top 10 results is trivially easy; an LLM can do it in five seconds. The only content that beats the LLM at synthesis is content that has something the LLM can't synthesize: new facts. Specific ways to generate Information Gain:

Run your own experiments. Test 15 tools, time each one, document results. Raw data beats summarization every time.
Use first-hand case studies. "A client we worked with saw X after doing Y" contains information that doesn't exist anywhere else.
Survey your audience. Your email list's aggregated answers to a question are genuinely new data.
Include calculations and tools. A working calculator that produces custom answers is more valuable than a page describing the calculation.
Interview practitioners. A quote from someone who actually does the thing trumps a paragraph of abstract explanation.
Show your work. Screenshots, process steps, real code snippets, actual outputs. Abstract explanations lose to concrete demonstrations.

A simple litmus test: if an LLM could write your article in ten seconds given only the top 10 search results, your Information Gain is zero. Write something the LLM couldn't produce without you.

5. Topical Authority

AI Overviews prefer citing sites that have demonstrable depth in the topic. A single "what is Kubernetes" article on a general marketing blog will rarely beat a dedicated devops publication with 200 Kubernetes articles over five years.

Topical authority is built by:

Publishing comprehensive coverage of a topic cluster, not scattered one-offs.
Linking related articles together into topic hubs (pillar + cluster architecture).
Consistent publishing cadence in the topic area.
Inbound links from topic-adjacent authorities.

If you're trying to rank for "cryptocurrency tax," you need a cluster of 10-20 related articles (capital gains, staking taxes, mining taxes, country-specific guides, tools), not a single mega-post.

Content Structure for LLM Retrieval

LLM retrieval works differently from traditional keyword matching. The AI is looking for semantic matches to the user's intent, not literal keyword matches. Content that optimizes for AIO looks different from content optimized for classic SEO:

Old SEO Practice	AIO-Optimized Practice
Repeat the exact-match keyword 15 times	Use varied phrasings of the concept
Long intros with "In today's world..."	Direct answer in the first paragraph
2,000 words to rank for anything	As many words as the topic genuinely needs
TOC and jump links for scannability	Self-contained sections that stand alone
Keyword stuffing in H2s	H2s as natural questions the user might ask
Thin FAQ section for schema	Substantive FAQ with unique answers
Generic meta descriptions	Meta descriptions that summarize the value prop

The shift: write for humans who might skip around, with a machine parser watching over their shoulder. Each section should make sense on its own. Each paragraph should have a point. Each sentence should add information.

Prompt Structure for AIO Citation

This sounds weird, but it works. Before writing a new article, draft three or four queries a user might type. For each query, imagine the AI Overview's sub-questions. Structure your article so each sub-question has a dedicated, well-answered section.

Example. Query: "best way to compress images for web." Likely AIO sub-questions:

What format should I use? (WebP vs AVIF vs JPEG)
What quality setting? (85 typical)
What tools can I use? (browser-based, CLI, Photoshop)
How much size reduction should I expect? (60-80% typical)
Will quality drop be visible? (usually not at 85)

Now give each sub-question its own H3 with a first-paragraph direct answer. The article still reads naturally for humans, but every likely AIO sub-question has a clean, citable passage. When the Overview engine decomposes the query, it finds your answers and cites them.

Measuring AIO Performance

Tracking AIO citations is harder than tracking rankings because Google doesn't report them directly. Your best proxies:

Search Console impressions without clicks — a rising "impressions" line with flat "clicks" often means you're being shown as a citation without driving traffic.
Branded search volume — users who see your brand in an AIO often search for you directly later. Track brand query growth as a leading indicator.
Third-party AIO monitors — SEO tools like Semrush, Ahrefs, and Surfer have added AIO citation tracking.
Manual spot checks — every week, search for your top 10 target queries in Incognito and see if/where you appear in AIO.

Also watch your clicks-per-impression ratio. If it's falling on queries where AIO is shown, you're being squeezed out. Where it's rising alongside flat rankings, you're benefiting from AIO citation visibility.

What Doesn't Work (and Wastes Your Time)

A quick list of 2026 SEO tactics that don't move AIO citation probability:

Keyword density optimization. The AI reads meaning, not word frequency.
Thin schema spam. Adding FAQPage schema to a page with no real FAQs gets flagged and penalized.
AI-generated filler. Google can detect mechanical LLM patterns. Pure LLM output scores low on Information Gain by definition.
"Skyscraper" tactics. Making a longer version of a competitor's post doesn't add Information Gain. More words isn't the same as more information.
Link farms and PBNs. These never worked reliably and they work even less now.
Gaming dates (bumping "Last updated" without editing). Google tracks actual content changes, not just timestamps.

A Practical AIO Optimization Checklist

Run this on any page you want cited in AI Overviews:

[ ] First paragraph of each H2 section is a standalone direct answer
[ ] FAQPage schema deployed with substantive Q&A (validated in Rich Results Test)
[ ] Article schema with author, datePublished, dateModified, publisher
[ ] Author bio page linked from byline; bio includes credentials and sameAs links
[ ] At least one unique insight, data point, or angle not in the top 10
[ ] Outbound links to 2-3 authoritative primary sources
[ ] Sections structured around likely AIO sub-questions
[ ] Meta description summarizes the value proposition, not just keywords
[ ] Core Web Vitals passing (INP ≤ 150ms, LCP ≤ 2.0s, CLS ≤ 0.1)
[ ] Clean breadcrumbs with BreadcrumbList schema
[ ] Visible "Last updated" date that reflects real edits
[ ] Internal links to related articles (topic hub pattern)

Tools That Help

Schema markup generator for valid JSON-LD across Article, FAQ, HowTo, and Product types.
Meta tag generator for Open Graph, Twitter Cards, and canonical tags.
Keyword density checker — use this to confirm you're NOT over-optimizing. Natural density beats keyword stuffing.
XML sitemap generator for helping Google discover new content fast.
Page speed estimator for the CWV layer of AIO ranking.
URL slug generator for clean, semantic URLs that help LLMs parse page intent.

The Tactical Roadmap

If you're starting from scratch, this is the order of operations:

Week 1: Audit your top 20 pages (by impressions in Search Console). Identify which ones are losing traffic to AIO.
Week 2-3: Rewrite each page's first-paragraph answers in each section. Add FAQPage schema with substantive Q&A.
Week 4: Audit E-E-A-T signals. Add author bios. Link sameAs profiles. Fix "About" page.
Week 5-6: Add Information Gain. Run an experiment. Survey your audience. Include original data.
Week 7-8: Technical SEO pass. CWV, internal linking, breadcrumbs.
Ongoing: Monitor AIO citation status weekly. Update content when rankings slip. Add new Information Gain periodically.

Frequently Asked Questions

What is Google AI Overview (AIO)?

AI Overview is the AI-generated summary Google shows above search results for many queries. It synthesizes an answer from multiple sources and cites 3-8 pages. AIO appears for roughly 60% of informational queries as of April 2026.

Do AI Overviews hurt traffic?

Yes, for informational queries. Sites cited inside AIO see their CTR drop by about 36% compared to pre-AIO position 1 click rates. Sites not cited lose 60%+. Sites cited but not ranking top 10 gain 200%+ — being cited beats ranking alone.

How do I get cited in AI Overviews?

Structure content as direct question-answer pairs, deploy FAQPage and Article schema, demonstrate E-E-A-T (author bios, credentials, primary sources), add Information Gain (original research, unique insights), and ensure Core Web Vitals pass the 2026 thresholds.

What is Information Gain?

Google's measure of how much new information your page adds beyond what's already in the index. Pages synthesizing existing content score low. Pages with original research, first-hand data, or unique perspectives score high. It's become a meaningful ranking factor in the AI era.

Is traditional SEO dead?

No. Links, technical SEO, CWV, and topical authority still matter. What changed is that content depth and Information Gain now carry more weight than keyword matching. The fundamentals that made great content in 2015 still apply; the cheap tactics that boosted mediocre content don't.

Should I block GPTBot and Googlebot-AI?

Blocking Googlebot-AI prevents your content from appearing in AIO. Most sites want the citation visibility and should allow it. Blocking GPTBot/ClaudeBot/PerplexityBot is a separate decision and depends on whether you want your content used to train third-party models versus appearing in their answers.