Chapter 1
SEO in 2027: What Has Changed and What Never Did
In two years, search has changed more than in the previous ten. AI Overviews (AI summaries above Google results) run in 200+ countries in 40+ languages, and in the summer of 2025 they had more than 2 billion monthly users. The conversational AI Mode has been available worldwide since October 2025, and in May 2026 Google reported more than a billion monthly users. Alongside them are AI assistants: search in ChatGPT (since late 2024), Gemini (a billion users, according to Alphabet’s third-quarter 2026 results), Claude with web search on all plans, Perplexity, and Microsoft Copilot. Classic SEO remains the foundation: Google’s AI answers are built on the same index and ranking systems. GEO goes on top of it.
What is gone. Keyword stuffing and bought links are named outright as spam in Google’s policies, and the SpamBrain algorithm zeroes out the weight of such links. In March 2024, three spam policies appeared: scaled content abuse (mass-produced pages made for rankings, whether a human or an AI writes them), site reputation abuse (“parasitic” hosting of third-party content on an authoritative domain; in November 2024 Google clarified that the site owner’s involvement does not protect against penalties), and expired domain abuse (expired domains used for junk content). At the same time, the “helpful content” system became part of core ranking. Then came core updates in August, November, and December 2024, March, June, and December 2025, and March and May 2026, plus spam updates, the latest in September 2026. Each time, sites that grew volume for volume’s sake lose rankings.
What works better than ever. E-E-A-T: in December 2022, Google added Experience, the author’s personal experience, to its guidelines for quality raters; Google does not call E-E-A-T a separate algorithmic factor, it is a quality benchmark. Primary sources: your own data, research, and case studies. Semantic clusters that cover a topic completely. Core Web Vitals: LCP, INP (which replaced FID in March 2024), and CLS are part of the page experience assessment; at equal relevance a fast site wins, but speed will not save weak text. Structured data (JSON-LD) helps search engines understand a page and earns rich snippets; Google does not require any special markup “for AI.”
What GEO is. The term Generative Engine Optimization was introduced by researchers from Princeton and other institutions in a 2023 paper (KDD 2024): in their experiments, edits to a text raised a source’s visibility in generative answers by up to 40%. The assistant finds pages, assembles an answer, and links to some of them; the goal of GEO is to land in those links and mentions. In its guide to generative search (May 2026), Google writes that SEO best practices remain the foundation for its AI features. For outside assistants, you add access for their crawlers, brand mentions in independent sources, and text that is easy to quote.
Three layers of search traffic. Classic Google and Bing links. Google’s AI features: their clicks and impressions are included in the Search Console Performance report, and since August 31, 2026 a separate generative AI report has been available to everyone (impressions only for now). Referrals from assistants, which show up in GA4 as referral traffic. Clicks from informational queries are falling: according to Pew Research (March 2025, 900 US adults), when an AI summary was shown, users clicked a regular result in 8% of visits versus 15% without it, and a link inside the summary in 1%. Ahrefs (300K queries, December 2023 versus December 2025, desktop) found the click-through rate of the first position 58% lower than expected where an AI Overview is present. Google, for its part, talks about billions of clicks a week from its AI features. The takeaway: the top of the funnel brings fewer visits, so count leads and signups, and work on getting named in the answers.
A realistic timeframe (a rule of thumb from practice). Months 1–3: the technical foundation, the first 15–20 pieces of content, the first impressions. Months 4–6: content expansion, the first links and mentions, top-10 positions for long-tail queries, the first inquiries. Months 7–12: the key clusters are covered, the brand appears in AI answers for niche questions, and organic search delivers a steady flow of leads. Traffic volume depends on the niche, so build your plan from a target number of inquiries.
Tools. Google Search Console and Bing Webmaster Tools (free; since February 2026 Bing has an AI Performance report on citations in Copilot). Google Analytics 4. PageSpeed Insights and Lighthouse. Rich Results Test and Schema Markup Validator. Ahrefs or Semrush (paid): competitors, links, queries, and AI answer visibility modules; for your own sites there is the free Ahrefs Webmaster Tools.
Chapter 2
Technical SEO: The Foundation Without Which Nothing Works
Content and links are useless if a search engine cannot index your site properly. Set up the technical foundation correctly once, then watch for regressions after releases. For AI search, Google’s requirements are the same: the page must be indexed and eligible to show a snippet, and there are no other technical conditions.
Crawling and indexing. Robots.txt must not block important sections; a common SaaS mistake is a blocked /blog/ or /pricing/ after a move from staging. Submit a sitemap.xml with all indexable pages to Search Console and Bing Webmaster Tools. Canonical: every page points to its main version through <link rel="canonical">. Look for problems in the Indexing → Pages report (formerly Coverage).
AI crawlers: whom to let in. Robots.txt rules are set for each bot separately. Google: AI Overviews and AI Mode run on the regular Googlebot; Google-Extended controls the use of content for training and for answers in other Google systems and does not affect search. OpenAI: OAI-SearchBot—display in ChatGPT search, GPTBot—training, ChatGPT-User—fetching at a user’s request. Anthropic: Claude-SearchBot—the search index, Claude-User—fetching at a user’s request, ClaudeBot—training. Perplexity: PerplexityBot indexes, and Perplexity-User fetches pages on request and, according to Perplexity’s documentation, usually ignores robots.txt. Copilot relies on the Bing index. A sensible setup: let the search bots in, and decide on the training bots separately. Check your CDN: since July 2025, Cloudflare blocks AI crawlers by default, so your site may be closed to assistants by a setting its owner does not know about. Check the logs for visits from OAI-SearchBot, Claude-SearchBot, and PerplexityBot.
JavaScript. Googlebot executes JavaScript, but according to a study by Vercel and MERJ (December 2024), the major AI crawlers (the OpenAI bots, ClaudeBot, PerplexityBot) see only the source HTML. If a SaaS product’s description, pricing, and documentation are assembled on the client, assistants do not see them. The solution is server-side rendering or prerendering of key pages (Next.js, Nuxt, Astro).
Core Web Vitals are measured on real users at the 75th percentile. LCP—loading of the main content, target under 2.5 s. INP—responsiveness to all interactions, target under 200 ms; on March 12, 2024, INP replaced FID. CLS—layout stability, target under 0.1. What helps: width/height on images, lazy loading below the first screen, modern image formats, less third-party JavaScript (chat and analytics widgets often hurt INP), and a CDN.
HTTPS is mandatory, and a free certificate is available from Let’s Encrypt and Cloudflare. Google indexes a site as a smartphone, so the content of the mobile and desktop versions must match. The Mobile-Friendly Test was retired in December 2023; check with Lighthouse and PageSpeed Insights.
Internal links distribute weight and show what matters most. Every important page gets links from several related pieces, and there should be no “orphans.” Breadcrumbs provide navigation and BreadcrumbList markup. Structure: cluster articles → pillar page → home page and product pages.
Structured data (JSON-LD). The basic set: Organization, Article, Product or SoftwareApplication, LocalBusiness for physical locations, BreadcrumbList, and Review/AggregateRating where reviews appear on the page. What changed: the sitelinks search box in the snippet (SearchAction) was removed in November 2024, HowTo rich results in 2023, and FAQ snippets have not been shown since May 7, 2026. FAQPage markup does no harm, but it no longer produces snippets. There is no special schema.org type for Google’s AI features. The main rule: the JSON-LD matches the visible text.
Hreflang for multilingual sites tells Google which language version to show to whom; without it, a search engine may show the wrong version or merge them.
404s and redirects. When you delete or move pages, set a 301 redirect to the closest matching URL, or links to the old address lose their value. AI crawlers request nonexistent URLs especially often, so an up-to-date sitemap matters twice over.
Speed audit. Run PageSpeed Insights for the mobile and desktop versions: field data at the top, a lab test with recommendations below. LCP is usually fixed by removing render-blocking resources and optimizing the main image, and INP by taking heavy scripts off the main thread.
Chapter 3
Keywords and Content Strategy
The goal of a content strategy is to build the most useful resource on a specific question in your niche. In its guide to generative search (May 2026), Google advises creating material based on your own knowledge of the topic rather than retelling what has already been written.
Finding queries: look for the questions your customers ask. Queries have gotten longer: according to Alphabet, queries in AI Mode are on average three times longer than regular ones, and one in five there is a voice query, a photo query, or one with a document. Sources: Ahrefs or Semrush (the Questions filter), AnswerThePublic, Google’s “People also ask,” Reddit and forums, support tickets and sales calls (the best raw material), and Search Console with a regex filter for “how,” “what,” “why,” “how much,” and “vs.”
Three types of queries. Informational (“how,” “what,” “why”) are the top of the funnel, for the blog and knowledge base; they are the ones AI answers most often cover, so the value of such articles now lies in being quotable. Navigational (a brand or product name) means people are looking for you specifically. Transactional and commercial (“buy,” “price,” “pricing,” “alternative to X,” “X vs Y”) are ready to act; these get product, service, pricing, and comparison pages.
Balancing difficulty and volume. A new site should start with low-difficulty queries (KD in Ahrefs or Semrush up to 20–30) and a specific intent. A narrow “DUNS number for Apple Developer” will bring fewer people than the general “DUNS number,” but first place for it is realistic and the visitor is exactly your audience; eleventh place for the general query yields almost nothing.
The content cluster (the Pillar + Cluster model). The pillar page is a detailed guide to a broad topic, for example “The Complete Guide to Business Credit.” The cluster articles are 15–25 pieces on narrow questions around it: “How to Get a Paydex of 80,” “A List of Net-30 Vendors,” “Quill vs. Uline for Business Credit.” Linking them to each other shows the search engine that the site covers the topic in full.
Content length. Google has no preferred word count, as its helpful content guidance says. The right length is as long as it takes to give a complete answer. Eight hundred words to the point beat three thousand words of filler: filler hides the answer from both the reader and the systems that look for something to quote.
E-E-A-T signals in content. An author with a bio, a photo, and verified qualifications on every article, plus an author page. Primary data: surveys, your own measurements, anonymized product data, personal experience. Publication and update dates. Links to primary sources. Real case studies with specific numbers. Screenshots and videos from your product: in its AI search guide, Google specifically advises backing up text with images and video where appropriate.
AI in content production. Google evaluates the quality of the result, and the line falls at the scaled content abuse policy: hundreds of templated pages made for rankings are a violation, no matter who wrote them. A safe setup: AI produces a draft, an expert adds experience and data, and an editor checks the facts.
A 6-month content plan. 1–3 pillar pages on key topics. 1–2 cluster articles a week. Pages for commercial queries: pricing, comparisons with competitors, and pages for specific use cases and industries. A monthly content audit: update outdated articles. Once a quarter, analyze search results and AI answers for your main queries: who is in the top 10, whom AI Overviews, ChatGPT, and Perplexity cite, and what your pages are missing.
Chapter 4
GEO: Getting Into AI Search Answers
GEO (Generative Engine Optimization) is work at the junction of SEO, PR, and the technical setup of your site. The goal: to have your content and brand named and cited by Google AI Overviews and AI Mode, ChatGPT, Gemini, Perplexity, Claude, and Copilot.
Why now. AI Overviews have more than 2 billion monthly users, AI Mode more than a billion, the Gemini app has reached a billion, and ChatGPT’s audience is counted in hundreds of millions per week. Some questions get settled right in the answer. If an assistant recommends three services and yours is not among them, you have lost the customer before they ever see the search results.
Where sources come from. AI Overviews and AI Mode find pages in the regular Google index and rely on its ranking systems. ChatGPT searches through its own crawler, OAI-SearchBot; a site closed to it does not appear in ChatGPT search answers. Perplexity has its own index (PerplexityBot), Claude uses Claude-SearchBot, and Copilot uses the Bing index. The first rule of GEO: all of these bots must be able to read your key pages (Chapter 2).
What raises your chance of appearing in an answer. Mentions: in an Ahrefs study of 75K brands, visibility in AI Overviews correlated most strongly with brand mentions across the web (0.664), and less with Domain Rating (0.326) and the number of backlinks (0.218). That is a correlation, but the conclusion is clear: other people have to write about you. Google also warns that scattering artificial “mentions” around is pointless. Specifics: in the experiments by the authors of the GEO paper, expert quotes, statistics, and source citations worked best, while keyword stuffing reduced visibility. Clear structure, fresh data, and uniqueness: Google advises making content that a dozen competitors do not have.
llms.txt: the real status. The format was proposed by Jeremy Howard (Answer.AI) in September 2024: a markdown file at /llms.txt with a description of the site and links to the main materials for AI agents. It does not control access; robots.txt does that. In June 2026, Google stated in its documentation that Google Search does not use llms.txt and the file does not affect ranking; special “AI files” for AI Overviews and AI Mode are not needed, according to Google’s guidance. OpenAI, Anthropic, and Perplexity, in their instructions for site owners, talk about robots.txt and their own bots; they have not publicly said that their search reads llms.txt. It is useful for API documentation and developer SaaS: the file is published, for example, by Anthropic’s documentation and many developer tools, and a developer can hand it to their own AI assistant. Structure: “# Company name,” “> Short description,” and a “## Key materials” section with links in the form “- [Title](/url): what is inside.” An hour of work, with no expectation of traffic growth.
Markup and FAQ. Google requires no special markup for AI answers, and FAQ snippets have not been shown since May 2026. The “question—short answer” format remains useful as visible text: 3–5 real customer questions with direct answers, phrased the way people ask in search and in chats.
Quotability. AI is more willing to cite text with specific numbers, an author, a date, and a source for the data, on a domain known in the topic. Phrase every claim you want to see in answers as a ready-made quote: one sentence, a fact, a source. Structure: an H2 as the question, the first 1–2 sentences as the direct answer, then details. There is no need to chop text into small fragments specifically for AI; Google says so directly.
Your brand in AI systems. For a question about a DUNS number, ChatGPT cites D&B, official sites, and major directories; your task is to get into that circle. Publish primary data and benchmarks, earn mentions in industry media and “best tools for X” roundups, and keep your profiles on G2 and Capterra (SaaS) or in Google Business Profile (service businesses) up to date: Google specifically notes that a business profile helps visibility in AI answers too. Honest “X vs Y” comparison pages give an assistant ready material for questions about choosing.
Monitoring. Manually: once a month, ask 20–30 questions on your topic in ChatGPT, Gemini, Perplexity, Claude, Copilot, and AI Mode—who gets named and who gets linked. Free: the generative AI report in Search Console (impressions in AI Overviews and AI Mode) and AI Performance in Bing Webmaster Tools (citations in Copilot). Paid: Ahrefs Brand Radar, Semrush AI Visibility Toolkit, Profound. According to Google, third-party tools have no access to its internal systems, so these are samples. If you are absent from the answers, work on authority and mentions.
Managing participation. Since summer 2026, Search Console has a switch that excludes a site from Google’s generative search features. A site that opts out loses impressions and traffic from them; according to Google, regular ranking is not affected. A business that needs customers almost never has a reason to opt out; you can hide specific fragments with nosnippet and data-nosnippet.
Chapter 5
Backlinks and Domain Authority
Backlinks remain an important ranking signal, and in the era of AI answers, brand mentions now count alongside them. One link from an authoritative, topically relevant site is worth more than a hundred from junk directories. SpamBrain zeroes out the weight of bought and exchanged links, and since 2024 the site reputation abuse policy has hit the “parasitic” sections of large media sites where placements were sold.
Links and mentions. In an Ahrefs study of 75K brands, visibility in AI Overviews correlated most strongly with brand mentions across the web, and noticeably less with the number of backlinks. A link works for rankings; a mention without a link helps assistants know about you and name you. Plan for both metrics.
What a good link is. One from a relevant site that is topically related to yours. One from a site with a real audience and traffic (Domain Rating in Ahrefs or Authority Score in Semrush is a useful but rough guide). With a natural anchor: the brand name, a URL, a descriptive phrase, without a run of exact-match keywords. Editorial: the author placed it themselves because the material is useful. Under Google’s rules, paid placements must be marked rel="sponsored" or nofollow; such links give you reach and a mention, but you should not count on them passing weight for ranking.
White-hat ways to earn links today.
Digital PR: content people want to quote. Original research (even a survey of 100 customers is primary data), annual benchmarks (“X Statistics in 2026” with your methodology), free calculators, and charts built on your data. Content like this collects links and media mentions on its own.
Journalist request services. Journalists look for experts to comment: you answer a request on your topic, get a quote, and often a link. The classic HARO has survived a change of owners: Cision shut down its successor Connectively in December 2024, and in April 2025 sold the HARO brand to Featured, which relaunched the newsletter. Working platforms: HARO, Featured, Qwoted, Source of Sources. Respond quickly and specifically: a number, personal experience, one or two sentences. Many major publications put nofollow on external links; the value of such publications is the mention, the trust, and the traffic.
Guest posts on relevant sites. Write for publications and blogs in your niche; find them by searching “[your niche] write for us” or “[your niche] guest post.” Criteria: live editorial content, topical closeness, a real audience. Be careful with scale: Google directly classifies mass campaigns of guest articles with keyword anchors as link spam. Write less often and better, for the audience and the mention.
Link Intersect. In Ahrefs (Link Intersect) or Semrush (Backlink Gap), enter your competitors’ domains: the tool shows the sites that link to them but not to you. These are the most promising targets for outreach.
Roundups and review platforms. For SaaS: G2, Capterra, Product Hunt, topical directories, and “best tools for X” articles. Being present there works for both links and GEO. Ask assistants “what should I choose for X” on your topic, see which roundups and reviews they cite as sources, and work to get included in exactly those.
Resource pages. Find “Useful resources” pages with searches like “[topic] + resources,” and write the owner a short note on how your material helps their readers.
Broken link building. Find links on other sites that lead to vanished pages on a topic close to yours. Tell the owner about the broken link and offer your material as a replacement. Check your own old URLs too: if a deleted page is still linked to, set up a 301 redirect.
What to avoid at all costs. Buying links on marketplaces, in “packages” on freelance platforms, or in PBN networks—Google finds them, and at best the links are zeroed out, at worst you get a manual penalty. Link exchange schemes. Spam in comments and on forums. Turnkey placements in other people’s sections on large domains: that is a direct violation of the site reputation abuse policy, and both the host site and your results take the hit.
Pace of growth. For a new site, a few quality links and mentions a month is a good result. A sudden jump of hundreds of links in a week with no news hook looks manipulative. Natural growth following publications and PR gives stable rankings.
Chapter 6
Local SEO and Google Business Profile
For a service business with an office or on-site visits in a specific city, local search is often the main source of inquiries. For queries like “[service] + [city]” and “[service] near me,” Google shows the Local Pack: a map and three businesses. In its guide to generative search, Google specifically writes that a business profile helps visibility in AI answers too.
How Google ranks local results. According to the official help page, by three groups of factors: relevance (how well the profile matches the query), distance (how far the business is from the person searching), and prominence (how well known the business is: links to its site, mentions, the number and ratings of reviews). Google also says outright that you cannot buy or request a higher local position.
Google Business Profile (GBP) is a free and essential tool. Register your business at business.google.com and complete verification. Choose an exact primary category and several secondary ones. Fill out the profile completely: description, hours (including holidays), services, service areas, attributes, real photos, a phone number, and a link to a relevant page on your site. Publish updates.
NAP consistency (Name, Address, Phone). Your name, address, and phone number must match everywhere: GBP, your site, Bing Places, Apple Business Connect (Apple Maps), Yelp, Facebook, and industry and city directories. Discrepancies confuse both the search engine and the customer, and AI assistants that collect information about a company from different sources may serve an outdated phone number or address. To check and fix data across dozens of directories, there are paid services like BrightLocal and Moz Local.
Reviews are the key to the Local Pack. According to Google’s help page, more reviews and good ratings help local ranking. Strategy: ask every satisfied customer to leave a Google review right after the service. GBP itself provides a short link to the review form; add it to the post-order email, the invoice, your signature, or a QR code on a business card. The rules are strict: Google prohibits paying for reviews or giving discounts for them, asking for reviews only from satisfied customers, and discouraging unhappy customers. In the US, an FTC rule banning the purchase and faking of reviews, with fines, has been in force since October 2024. Violations end in removed reviews, profile restrictions, and trouble with the regulator.
Replies to reviews. Reply to every review: future customers see it. Respond to negative reviews calmly: acknowledge the problem, offer a solution, and move the conversation to a private channel. “Thank you for writing. Please contact us directly—we want to get to the bottom of the situation.”
Local queries. “[service] in [city],” “[service] [neighborhood],” “best [service] [city],” “[service] near me.” For each city or neighborhood where you really work, make a separate page with unique content: an address or service area, examples of work in that neighborhood, reviews from local customers, and prices. Google treats clone pages with only the city name swapped as doorway pages, which violates its spam policy.
Local links and mentions: the chamber of commerce, associations, city media, local directories, event sponsorships, and partnerships with neighbors. All of this works for prominence, one of the three local ranking factors.
Structured data for a local business. LocalBusiness markup (or a more precise subtype) on the home page and location pages: name, address, phone, opening hours, price range, photos, and coordinates. It helps Google match the site to the profile. Note Google’s rule: star ratings do not appear in the snippet for reviews that a business publishes about itself on its own site. The data in the markup must match GBP and the visible text.
Tracking local positions. Your position in the Local Pack depends on where the person is standing, so a single number means little. Local Falcon and BrightLocal show your position on a geographic grid—how you look from different points in the city—and help you find neighborhoods where a competitor beats you. Once a month, also ask ChatGPT, Gemini, and AI Mode “which [service] company should I pick in [city]” and record whether they name you and where they get the information.
Chapter 7
Analytics and Measuring SEO Results
“What doesn’t get measured doesn’t get improved.” SEO comes with plenty of data, and most of it is noise. With AI answers, impressions grow faster than clicks, and some of the influence does not show up in analytics at all. Here is what is worth measuring.
Google Search Console is the main tool: free, with data from Google itself. The Performance report: queries, pages, devices, countries. Impressions and clicks from AI Overviews and AI Mode are included in the overall “Web” totals and are not broken out there. The generative AI report (since August 31, 2026, for all sites with enough impressions): how many times links to your pages were shown in AI Overviews and AI Mode, by page, country, date, and device; it has no clicks yet. “Indexing → Pages” shows indexing problems. Core Web Vitals shows speed for real users.
Bing Webmaster Tools. Since February 2026 it has an AI Performance report: how many times your pages were cited in Microsoft Copilot and Bing’s AI summaries, and which “grounding queries” found them. There are no clicks, but it is a rare free source of citation data outside Google. Sites can be imported from Search Console in a couple of minutes.
Key metrics. Average position shows the trend. CTR is the share of clicks out of impressions: at position 3 with a CTR of about 1%, check the title and description, but first open the results page—if an AI Overview sits above you, a low CTR may be normal. A big gap between impressions and clicks gives you a list of pages whose snippets and quotability need work.
GA4. The Organic Search channel. Engagement rate and its inverse, bounce rate. Average engagement time. Key events (what GA4 has called conversions since 2024): which pages bring signups, inquiries, and payments. Set up traffic from assistants as its own channel group, by the sources chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, and copilot.microsoft.com: the volume is still small, and it is worth looking at conversion separately.
Ahrefs and Semrush. Domain Rating or Authority Score is a rough guide to authority. Organic Keywords in the top 10 and top 100. New and lost referring domains. Competitor Gap—competitors’ queries that you do not rank for. Brand Radar and AI Visibility Toolkit—the share of brand mentions in assistants’ answers, estimated on a sample of prompts.
The monthly report. Organic traffic against last month and the same month a year ago. New queries in the top 10. Impressions in the generative AI report and citations in Bing. A manual check of key questions in assistants. New referring domains and mentions. Technical errors. Conversions from organic search and assistants. Branded queries: growth in them without new ad campaigns may mean you have started to be mentioned in AI answers.
Traps. Looking at positions instead of inquiries: third place with no conversions is worse than eighth place that brings customers. Comparing traffic without accounting for seasonality—look year over year. Panicking over a drop in informational clicks without checking inquiries. Working only on new content when updating strong old pages often yields more.
Content decay. After a year or a year and a half, many articles lose rankings: the data goes stale and competitors write better. Once a quarter, find pages with falling clicks (compare periods in the Performance report), update the figures, prices, and screenshots, add sections and questions, strengthen internal links, and look at what the AI Overview answers for that query. Change the update date only after a substantive edit: Google advises against using dates to fake freshness.
SEO ROI. In the first months the return is negative, then organic search becomes one of the cheapest channels. A hypothetical SaaS example (an illustration of the calculation): $2,000 a month for a writer and tools; by month 12, 5,000 visitors a month, 2% start a trial (100), and 10% of them pay $100 each. Every month adds $1,000 to MRR, so, ignoring churn, in two to three months at this level recurring revenue covers the budget, and after that the gap grows. Plug in your own conversion rates and price.
Chapter 8
Monetizing Organic Traffic
Traffic does not pay the bills by itself. Organic search has to turn into inquiries, trials, and payments, and that matters more now than before: there are fewer informational visits, but the people who arrive after an AI answer often already know what they want.
Search intent. Informational: the person is getting to grips with a topic; the goal is contact through a lead magnet or a newsletter. Commercial: the person is comparing; the goal is comparisons, case studies, reviews, and a demo. Transactional: ready to pay; the goal is a pricing page and a clear call to action. In a conversation with an assistant, the same three intents appear: “how to do X,” “what is better for X,” “how much does X cost at Y.”
Lead magnets. A checklist or template that goes with an article. A free tool or calculator (it collects both contacts and links). An email course of 5–7 emails. A comparison table or an industry benchmark. The rule: the lead magnet is more valuable than the article itself; if people will not leave an email for it, it is not valuable enough.
Calls to action. In every article, a clear next step: at the beginning for those ready to act, after the main block of value, at the end, and as a pinned button. Replacing “Sign up” with “Start free” or “Demo in 2 minutes” can noticeably shift conversion; how much, an A/B test on your own traffic will show.
Product content. The best SEO content for SaaS shows the product in action: instead of “How to Track Working Time,” write “How to Track Working Time in [your product]” with screenshots and scenarios. Pages like these also answer assistant questions of the form “how to do X in Y.”
Comparison pages. “[Your product] vs. [Competitor]” and “[Competitor] alternatives” convert best: the person is already choosing. Write honestly, admit where the competitor is stronger, and add a summary table with prices, features, and the date of the last update. Start with the top 5 competitors.
Retargeting. Most visitors do not buy on the first visit. The Google Ads, Meta, and LinkedIn pixels let you bring them back with ads, usually cheaper than cold traffic. To a reader of the pricing page: “Start free.” To a reader of a comparison: your advantages. For visitors from the EEA, you need a consent banner and Google Consent Mode v2, or Google remarketing does not work on them.
Attribution. GA4 and UTM tags on all campaigns. Since 2023, GA4 has kept two attribution models, data-driven and last click; Google removed first click, linear, time decay, and position-based. For SaaS, add a “How did you hear about us?” question to the signup form with the options “Google search,” “ChatGPT or another AI assistant,” “Referral,” “Social media,” and a free-text field. A person often reads about you in ChatGPT and then types the name into Google or arrives directly; self-reporting and the trend in branded queries are the main way to see this channel.
The book’s main takeaway. A technical foundation, useful and unique content, and honest links and mentions have always worked and still work. What is new is that your reader is increasingly an AI system that decides whom to name to a person. It needs accessible HTML, specific facts, clear structure, and confirmation from independent sources. Give it that, and organic search will remain one of the cheapest ways for SaaS and service businesses to win customers.
Sources (verified in November 2026)
1. Google Search Central—AI features and your website (requirements for AI Overviews and AI Mode, reporting): https://developers.google.com/search/docs/appearance/ai-features
2. Google Search Central—Optimizing your website for generative AI features on Google Search (guide, May 2026): https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
3. Google Search Central—Latest Google Search documentation updates (FAQ snippets, llms.txt, structured data, spam policies): https://developers.google.com/search/updates
4. Google Search Central—Spam policies for Google web search (scaled content abuse, site reputation, expired domain abuse, link spam): https://developers.google.com/search/docs/essentials/spam-policies
5. Google Search Status Dashboard—history of core and spam updates, 2024–2026: https://status.search.google.com/products/rGHU1u87FJnkP6W2GwMi/history
6. Google—New opportunities, control and insights for website owners (the generative AI report and the switch in Search Console, June–August 2026): https://blog.google/products-and-platforms/products/search/new-controls-website-owners/
7. Alphabet—CEO remarks: second quarter of 2025 (AI Overviews, 2 billion users) https://blog.google/inside-google/message-ceo/alphabet-earnings-q2-2025/ and third quarter of 2026 (Gemini, AI Mode) https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q3-2026/
8. Google—Search at I/O 2026 (AI Mode, more than a billion monthly users): https://blog.google/products-and-platforms/products/search/search-io-2026/
9. web.dev—Interaction to Next Paint is officially a Core Web Vital: https://web.dev/blog/inp-cwv-launch
10. AI crawler documentation: OpenAI https://developers.openai.com/api/docs/bots, Anthropic https://support.claude.com/en/articles/8896518, Perplexity https://docs.perplexity.ai/guides/bots
11. Microsoft Bing—Introducing AI Performance in Bing Webmaster Tools (February 2026): https://blogs.bing.com/webmaster/February-2026/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview
12. Pew Research Center—Google users are less likely to click on links when an AI summary appears (July 2025): https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
13. Ahrefs—the effect of AI Overviews on CTR (2026 update) https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/ and factors in brand visibility in AI Overviews (75K brands) https://ahrefs.com/blog/ai-overview-brand-correlation/
14. Aggarwal et al.—GEO: Generative Engine Optimization (KDD 2024): https://arxiv.org/abs/2311.09735; Vercel—The rise of the AI crawler (December 2024): https://vercel.com/blog/the-rise-of-the-ai-crawler
15. llms.txt—the original format proposal: https://llmstxt.org/; Google Business Profile—how Google ranks local results: https://support.google.com/business/answer/7091
Download this guide in EPUB and PDF
