How to Get Cited in Google AI Overview in 2026

For context: from 2014-2022, Google had the "featured snippet" — an answer box at position zero displaying a direct quote from a single source. If your content made it into the featured snippet, one click from you replaced 10 clicks from a normal position 1. That was the main SEO target in Indonesia for years. Starting in 2024-2026, Google replaced that pattern with AI Overview, which is fundamentally different: (1) The answer is no longer a verbatim quote, but a summary rewritten by Google's AI from multiple sources, (2) The sources referenced aren't just one, but usually 3-7 websites that contribute to the answer, (3) The AI can "combine" facts from several sources at once, then tie them to brand mentions or reviews. The implication for Indonesian small businesses: the way content is read by the system has changed. Content that used to suit the featured snippet (a 50-60 word paragraph that answers the question directly) is still relevant, but now it has to be supplemented with a more "modular" structure — information organized into small units (passage-level) that the AI can pick up as a citation. Your content no longer competes for "who shows up at position 1" but for "who gets mentioned in the AI's answer paragraph."

One of the most important principles in the AI Overview era is content that's "passage-level optimizable" — meaning each paragraph can stand on its own as an independent unit of information. This concept runs counter to traditional blogs with lots of "transitions" and paragraphs that only make sense when read together with the previous one. The practical way to apply it: (1) Start each paragraph with a clear topic sentence — if that paragraph is taken in isolation, the reader still understands the context. (2) Avoid "cliffhangers" between paragraphs like "and for the next part..." or "let's take a look...". AI prefers content that's to the point. (3) Use H2/H3 headings phrased as a question or a specific claim — not a poetic title. For example: "What is the average price of a homestay in Sarangan?" is better than "Exploring Sarangan Accommodation." (4) Provide a direct answer in the first 40-60 words of each section, then expand with detail. AI Overview likes a "self-contained answer" at the start of a section. For Webiti clients, we refactored existing blogs from a long-narrative format into a modular Q&A format — the consistent result: visibility in AI Overview rose 3-5x within 60-90 days. No miracle, just an approach that matches how machines now read.

Schema.org markup (JSON-LD) is how a website "tells" Google and AI what the page contains, with a formal structure. For the AI Overview era, schema is no longer optional — it's the data-exchange language between a website and AI systems. Indonesian small businesses need at least 3 types of schema depending on their industry:

• LocalBusiness (or its subtypes: Restaurant, MedicalClinic, BeautySalon) with full address, opening hours, geo coordinates, service area, price range, and aggregate rating. Critical for appearing in AI Overviews that answer local questions like "nearest dental clinic in Madiun."
• FAQPage for pages containing questions and answers — AI Overview often pulls answers from FAQs whose schema is marked up. Make sure each Q&A is a real question customers frequently ask, not a filler question.
• Article for blog content and tutorials. Include this metadata: author with full name + bio, datePublished, dateModified, hero image. AI Overview trusts articles with clear author identification more. (Note: the HowTo rich result was retired by Google back in 2023, so there's no longer any need to install it for the purpose of appearing in search results.)
• BreadcrumbList to help AI understand the page hierarchy.
• Product or Service for sales pages, with price, availability, and rating.

What we often see from cheap vendors: schema is installed but with the wrong fields — this can be worse than having no schema at all, because Google marks it as "untrustworthy structured data." Check it with the Google Rich Results Test before feeling safe.

AI Overview doesn't only read the content on your own website — the system also evaluates how often your brand is mentioned across other authoritative sources on the internet. This is called the "brand mention signal" or "unlinked citation." For Indonesian small businesses, some sources the AI considers authoritative: (a) Local/national online media (detik.com, kompas.com, jawapos.com, regency media), (b) Google Maps reviews and replies, (c) Wikipedia (if the brand is already big enough), (d) Forums and community groups (Kaskus, Quora ID, Reddit Indonesia), (e) Verified business directories. How to build mentions ethically: (1) Organic PR — pitch your business story to local journalists who write about food or small businesses. Not a paid press release, but a story that's genuinely interesting. (2) Be active on Google Business Profile with weekly posts, respond to all reviews (good and bad), upload photos regularly. (3) Sponsor local community events, which usually generates local media coverage. (4) Be a featured guest on a relevant blog or podcast. What NOT to do: buy backlinks on PBN networks, buy fake reviews, or do "mass directory submission" that's already considered spam. The AI of 2026 is sophisticated enough to tell organic mentions from manipulative ones. What you want to avoid: having your brand associated with manipulative signals — the AI might actually steer your brand away from results it considers high quality.

Starting in 2024, many large sites began installing an `llms.txt` file (similar to `robots.txt` but for AI crawlers). This file tells AI systems how to read your website correctly — which content is authoritative and which is secondary. For a small business, the `llms.txt` file can contain: a short brand summary (1-2 paragraphs), a list of the main pages you want AI to cite with a short description for each, and a priority ranking of pages. A simple example for a pecel stall: a summary like "Pecel Bu Sari is a traditional floor-seating eatery in Madiun since 1987, specializing in pecel with a family-recipe peanut sauce," then a list of pages: Menu (link), Our Story (link), Group Reservations (link), Location & Opening Hours (link). This file is hosted at the root domain (businessname.com/llms.txt) and is read by AI crawlers during indexing. Not all AI uses it yet — it's still an evolving standard — but some systems (Perplexity, some ChatGPT search features) have started prioritizing sites with a good llms.txt. The implementation is very lightweight (a plain text file under 5KB) and there's no downside. Vendors who are serious about AI readiness have already started installing it for all clients. If your vendor hasn't, ask them to add it — it's free and future-proof.

Internal linking — links between pages on the same website — now has two functions: for user navigation (long understood) and for helping AI understand the "topology" of your brand (the newer one). The AI of 2026 uses the internal-link structure to determine which page is the most authoritative for a given topic on your domain. If page X is linked from 8 other pages with relevant anchor text, the AI concludes that page X is the "pillar page" for that topic. A practical strategy for small businesses: (1) Identify 3-5 "pillar pages" on the website — the most important pages you want to rank highly. For a clinic: Online Booking, Lead Doctor Profile, Flagship Services. (2) Make sure every other page on the site has at least 1 relevant link to a pillar page, with natural anchor text (not "click here" but "Klinik Permata's dermatology specialist"). (3) The blog or educational content is a multiplier hub — every blog article should link to 2-3 relevant service/product pages. (4) Avoid over-linking — if every word on a page becomes a link, the AI considers it spam. Rule of thumb: 1 link per 100-150 words is natural. (5) Check for orphan pages — pages that aren't linked from anywhere on the website. Such pages are considered low priority by AI. Audit your internal-link structure at least every 6 months, especially after adding new pages.

After implementing the 5 pillars above, how do you check whether it's effective? AI Overview doesn't have an official dashboard from Google (not yet, as of 2026). But there are approximate methods we use for clients: (1) Manual query check — search your business's main keywords on Google in incognito mode (so you're not biased by your personal profile). Does the AI Overview appear? Is one of the cited sources your website? Do this every 2 weeks for 10-15 main keywords. (2) Search Console — even though Google hasn't yet given a specific "AI Overview impressions" metric, there's an interesting pattern: if impressions rise but CTR falls for certain keywords, your page likely appears often but users already get the answer in the AI Overview without clicking. This isn't entirely bad news — it means you're getting brand exposure without the click. Visibility is still valuable. (3) Direct mention test — ask ChatGPT, Perplexity, or Gemini a question about your industry in your location, and see whether your brand gets mentioned. For example: "best dental clinic in Kediri" or "cheap homestay near Lake Sarangan." If your brand consistently appears in these 3 AI systems, you're already "AI-discoverable." (4) Track referral traffic from AI sources in Google Analytics — domains like `chatgpt.com`, `perplexity.ai`, and `gemini.google.com` are starting to show up in referral traffic for sites that are already well optimized.

For transparency, there are a few things people fuss about online regarding AI Overview that, in our field experience, aren't yet critical for Indonesian small businesses in 2026: (1) "AI-generated content vs. human content" — Google has officially stated it doesn't discriminate against content based on whether it was made by AI or a human, as long as it's high quality and helpful. If you use AI for an initial draft and then have a human edit it with local context, that's fine. What gets penalized is "mass-produced AI spam" with no added value. (2) "Optimizing specifically for ChatGPT" — some consultants sell "GPT optimization" services at high prices. The fact is: the principles that make content good for Google's AI Overview generally also make that content good for all AI (ChatGPT search, Perplexity, etc.). There's no need for a separate strategy per platform. (3) "Domain authority is dead" — DA/DR from Moz/Ahrefs is indeed not a Google metric, but signals of trust and backlink quality are still very relevant for AI Overview. Focus on quality, not the number. What to remember: AI search is still very new in Indonesia. Indonesian isn't yet as well handled as English in AI indexing. But the pattern is starting to show, and the early-mover advantage (small businesses that are ready as early as 2026) will become apparent in the next 12-18 months.