How to Do SEO in 2026: The Small-Business Guide That Still Works When Google Answers Without a Click

What changed, in numbers

Three facts set the 2026 playbook. In the fourth quarter of 2025, 56% of Google desktop searches ended with no click, and more on phones, according to Digital Applied's analysis of click data. Google's AI Overviews appear on roughly one in ten US searches and are being extended to local and transactional queries through 2026 and 2027. And ads inside AI Overviews went from about 3% of those result pages in January 2025 to about 40% by November 2025. The page that ranks first still matters, but the answer box above it is where a growing share of customers stop.

None of this makes SEO optional. It changes what SEO is for. A business that is the source of the AI answer gets the call whether or not anyone clicks. A business that only ranks in the list underneath gets what is left.

Step 1: One page per service and per city

Google and AI systems answer specific questions with specific pages. A homepage that lists everything you do ranks for nothing in particular and gives an AI nothing to quote. A plumber with pages for water heaters, repiping, drain cleaning and slab leaks, each written for Escondido, Vista and San Marcos, has thirty pages that can each be the answer to a real question. This is the single biggest lever for a local business, and the one most sites skip.

• Write each page for its own question, with its own prices, examples and FAQs. A city page that is a copy with the name swapped is a doorway page and Google filters it.
• Put the city and the service in the title, the first heading and the first sentence.
• Link every page to the related service and city pages, so the site reads as one connected business.
• Demo Diego, a demolition company in San Diego, runs 300+ of these pages and gets calls from every corner of the county without ads.

Step 2: The answer in the first 100 words

Neuroflash's 2026 citation analysis found that 44% of ChatGPT citations come from the first 30% of a page, and that cited passages are almost twice as likely to use definitive language (36% against 20%) than hedged language. So every page opens with two or three plain sentences that answer the question, with the number in it: what it is, what it costs, how long it takes, who it is for. The rest of the page supports that answer. Write 'A water heater replacement in Escondido costs $1,800 to $3,200 and takes one day', not 'Contact us to learn more about our water heater services'.

Step 3: Structured data that names the business

Schema markup is the part of the page written for machines. In 2026 five types matter for a local business: LocalBusiness or ProfessionalService with name, phone, address or service area, hours and the same details as the Business Profile; Service for each service page; FAQPage for the questions on each page; BreadcrumbList; and Article or BlogPosting for guides. Speakable markup on the answer paragraph and the FAQ tells assistants which sentences to read aloud. Schema does not make a bad page rank. It makes a good page legible to the systems that assemble answers.

Step 4: A Google Business Profile that matches

The map block with three businesses that sits above the results comes from the Business Profile, and Google's stated direction is to bring AI Overviews to local queries next. The profile needs the right primary category, a service area or address that matches the site, hours that match, the services list filled in, ten or more real photos, and reviews that get answered. Then it needs to be linked from the site's structured data so Google treats the website and the profile as one entity. A profile with mismatched hours or a different phone number is a trust problem, and trust is what AI answers are built on.

Step 5: A site that loads in under a second

Speed is a tie-breaker in rankings and a filter in AI systems, which crawl fast pages more often and more completely. A hand-coded site on Next.js or plain HTML loads in under a second on a phone and scores 95 or higher on Google PageSpeed. A typical WordPress site with a theme and twenty plugins scores 40 to 60 and takes three to six seconds. This site is the demonstration: 149 pages, sub-second load, 99 to 100 on every Lighthouse category. Test any site free at pagespeed.web.dev before you believe a claim about it.

Step 6: Let the AI crawlers in

A page cannot be cited by a system that is not allowed to read it. GPTBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot, Google-Extended (Gemini training) and Bingbot each check robots.txt, and a surprising number of business sites block them by default through a plugin or a host setting. Allow them explicitly. Publish an llms.txt file that lists your most important pages in plain text. Keep a sitemap that is current and submitted in Search Console, and request indexing for new pages instead of waiting.

Step 7: Get mentioned where AI looks

This is the step that did not exist in 2020. Across the 2026 studies, brands are about 6.5 times more likely to be cited by AI through third-party sources than through their own website, Reddit is the most-cited domain across ChatGPT, Gemini, Perplexity and AI Overviews, and LinkedIn is the most-cited domain for professional queries. For a local business that means: reviews on Google and Yelp with real detail in them, a listing on Clutch, Expertise, Thumbtack or the directory for your trade, one genuinely useful answer in the local subreddit when someone asks for a recommendation, and the owner posting real work on LinkedIn under their own name. Not spam. Presence.

What to stop doing

• Stop buying links. AI systems weigh mentions and reviews, not link counts, and Google's spam systems catch bought links faster every year.
• Stop publishing AI-generated filler. Ten thin articles a week get nothing cited and dilute the pages that could be.
• Stop measuring traffic alone. Track calls, forms, branded searches and citations. A page that answers the question inside AI Overviews and gets the call did its job with zero clicks.
• Stop guaranteeing rankings, or believing anyone who does. Google and the Federal Trade Commission both act on that claim.

A 90-day plan for a local business

The order matters. Search and AI research first, then the page map, then design, then build. Design-first sites are the most common thing we replace: beautiful, five pages, invisible.

What this costs

A custom site built this way costs $3,000 to $8,000 for a business with a handful of services and one city, and $8,000 to $25,000 for a multi-city site with dozens of pages. After launch, care is $99 a month and growth plans that add new pages every month run $500 to $2,000. The math to check before spending any of it: one lead is worth your average job times your gross margin times your close rate. Divide the monthly plan by that number and you have the calls per month that break even. We publish the calculator on our pricing page.