Starting SEO on a brand-new website in 2026 means starting for two surfaces at once: the rankings your analytics can see, and the AI answers your analytics mostly can't. The good news is that one sequence of work feeds both, and a zero-authority domain gets to skip every bad habit older sites are busy unlearning. Here's the first 90 days, in the order that compounds, built from what practitioners actually tell founders asking this exact question and from what the engines now reward.
Days 1-7: Foundations and the Crawler Decision
Week one is plumbing, and it gates everything after. Make the site crawlable and fast enough to be boring: clean URLs, an XML sitemap, no staging noindex left behind, pages that load without drama. Verify Search Console and Bing Webmaster Tools so the data starts accruing from day zero. Google's SEO starter guide remains the canonical checklist for this layer, and it's free.
The 2026 addition is the crawler decision. Your robots.txt now governs two populations: search crawlers, which you obviously allow, and the AI roster, GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and the rest of the cast, which most new businesses should also allow, since AI answers are a discovery surface and young domains sometimes get cited there before they rank anywhere. Whatever you choose, choose it deliberately and check what your host or CDN defaults to, because silent edge blocks are the classic way a new site becomes invisible to engines it never decided to refuse.
Days 1-14: Make Your Entity Legible
Before content, identity. Engines that select trusted sources need to resolve what your company is, and a new domain starts with zero corroboration. Fix that mechanically: the same company name, description, category language, and pricing everywhere machines read, your site, LinkedIn, your Google Business Profile if you're local, the two or three review or directory listings that fit your category, and your social profiles. Add an llms.txt if you like, with honest expectations, since it's ten minutes and no more.
This step is unglamorous and it punches above its weight, because entity consistency is one of the few levers a brand-new site controls completely on day one. Nobody can out-authority you on the facts of your own company.
200+ SaaS teams already track their AI citations.
They know exactly when ChatGPT mentions their brand, and when it stops. Do you?
Days 7-30: Plan One Subject Instead of a Keyword List
Here's where most new sites lose the year before writing a word: they pull a keyword list and start publishing unrelated articles into it. The plan that works is subject-first. Pick the one subject your site should own, write down the page that will define it, then collect the real questions buyers ask about that subject and group them by search results, since two queries returning the same ten pages are one article. That grouped list is your first quarter's content, and the fate of every article in it gets decided by exactly this planning.
Filter the list through the zero-authority reality. A new domain wins long-tail, specific, low-competition questions first, especially comparison and how-to queries where the current answers are thin. Skip anything whose top ten is wall-to-wall high-authority domains; you'll come back for those in year two. One practitioner's first-90-days summary in a thread on this exact question got the order right: technical setup, one tight content cluster, and Search Console data, with link chasing deferred, because a five-page site has nothing worth linking to yet:
If you had to start SEO from scratch in 2026 with a brand-new website, what would you focus on during the first 3–6 months?
I keep seeing conflicting advice about backlinks, content volume, topical authority, and technical SEO. Curious what experienced SEOs here would prioritize first.
Days 30-90: Publish the Cluster in Order
Order matters more than volume. The defining page goes live first, or at least early, so every supporting article has a parent to strengthen. Then the supporting pieces, each answering one grouped question completely, each linking up to the definition page and across to two siblings. Write every section so it survives being quoted alone, direct answers under honest headings, because passage-level extraction is how AI engines read you, and a new site's first citations usually arrive passage-first.
Cadence honesty: a steady weekly rhythm beats a launch-month blitz that dies in week six. And calibrate month three against the real arc, impressions before clicks and long-tail before head terms, with a few hundred impressions marking the seedling stage, so you don't rebuild a healthy young site out of impatience.
Days 30-90: Earn the First Ten Links
Once the cluster's best pages exist, links start mattering, and the target is the best page rather than the homepage. Ten honest moves beat a hundred bought ones, and the buying route now carries more risk than reward. The motions that work at this size: answer questions in the communities where your category gets discussed, with disclosure and real answers. Pitch the writers who cover your niche with something specific, your data or your unusual angle, rather than "please link me." List the site in the directories and review platforms buyers actually check. And tell every partner, customer, and friendly vendor who'd genuinely mention you that the page exists. Ten links earned this way carry more weight than they look like they should, because on a young domain each early link is a meaningful share of the whole profile, and engines read the company you keep.
How often does ChatGPT mention your brand?
Most founders have no idea. The answer might surprise you.
A Worked Example: New B2B SaaS, Month One
Concrete beats abstract, so here's the plan shape for an imaginary invoicing tool launching today. The subject it should own: getting paid faster as a small agency. The defining page: a complete guide to agency payment terms, the page every supporting article will point at. The question list, grouped from real searches: how to write payment terms into a contract, what to do when a client pays late, deposit or milestone billing, net-30 versus net-15 for agencies, how to automate invoice follow-ups, and seven more of the same shape. Every one is long-tail, answerable completely, and absent a dominant incumbent answer.
Notice what's missing: "best invoicing software," volume 40,000, top ten all giants. That query goes on the year-two list. Month one publishes the defining page plus four supporting answers, each linking up and across, and the site's first citation realistically arrives on the net-30-versus-net-15 comparison, in an AI answer, before Google moves at all.
What To Skip in Year One
The anti-checklist, because beginner time leaks into these holes. Skip daily rank checking; weekly is signal, daily is noise on a new domain. Skip authority-score obsession; DA and DR are vendor metrics, and watching them wobble teaches nothing. Skip bought link batches entirely, at any price that seems attractive. Skip schema-everything plugins that markup pages nobody asked about; FAQ and Article schema on your cluster is plenty. And skip the tool-stack shopping spree; Search Console, one rank tracker, and a citation baseline cover the first six months, and every additional dashboard is a place to hide from publishing.
Day One: Run Both Scoreboards
The measurement habit separates sites that learn from sites that guess, and in 2026 the scoreboard has two halves. Rankings and Search Console cover the surface you can see. The other half is what AI engines say when someone asks about your category: whether you appear, who gets cited instead, how each engine describes you, and whether that description is even accurate. Take that baseline in week one, while the answer is probably "you don't appear anywhere," because the day-zero picture is what makes month-three progress visible and provable.
New-site owners are routinely surprised which scoreboard moves first. Comparison pages on young domains get pulled into AI answers with startling speed when they're the clearest source on a specific matchup, sometimes while the same page sits on Google's page four. We built our tracking to watch both halves weekly across ChatGPT, Perplexity, Claude, Gemini, and Google's AI surfaces, and RankControl runs the whole sequence in this guide, the planning, the cluster publishing, the outreach drafts, and the weekly checks, as one pipeline for teams that want the system rather than the checklist.
What the Old Courses Won't Tell You
Let me back up a step and address the question every beginner actually googles first: are the famous six-year-old SEO courses still worth taking? The community consensus is now roughly "fundamentals yes, worldview no," and one commenter put the gap perfectly: they're trying to get a small WordPress site mentioned in AI Overviews and Perplexity, and none of the old material even mentions that surface. The fundamentals those courses teach, keyword research, intent, on-page hygiene, internal linking, survive completely. What's new is the second surface and its physics: answer absorption changing which queries still click, entity trust deciding who gets cited, extraction-friendly structure paying twice, and a second roster of crawlers reading your robots.txt.
Google itself drew the line under this in May, publishing official guidance on optimizing for generative AI features and framing it as ordinary SEO. That's the whole 2026 story for a new site: the work didn't change shape, the scoreboard grew a second half, and a domain that starts today gets to build for both from its first page. Start the sequence, keep the rhythm, measure both halves, and let the compounding do what it does.
15 hours a week manually. Or 15 minutes with RankControl.
Track citations, monitor competitors, and fix content gaps across every AI search engine. Automatically.
