SEO for Startups: When to Start and What to Do First

1. Blog
2. Discoverability & Content

By Saroj Jha · October 5, 2026 · 15 min read

Start the technical basics now, because they take an afternoon. Write a few pages for people already searching for a fix early, because they convert. Build a content cluster once you know the words your buyers use. Judge it in months, not weeks, and judge the early months on what Search Console shows, not on traffic.

"Should we start SEO now?" is a question almost every early founder asks, usually followed by "how long does it take?" and "how many blog posts do we need?" The honest answer to the first is that SEO is not one decision. It is three jobs with different costs and different timing, and most early-stage mistakes come from treating them as one: either doing nothing because "SEO takes a year", or publishing fifty posts in a quarter and wondering why nothing happened.

Below are five steps in the order I would do them with a founder, each followed by one thing to do this week. This is AAJ's judgement, not a guarantee of results. It covers classic search. Being cited in AI answers has its own guide, the GEO and AEO Playbook, and the terms SEO, GEO and AEO are explained in this earlier post.

Running case: Example Co. Fictional

Example Co. is a fictional seed-stage AI tool that turns B2B sales-call recordings into CRM notes and follow-up emails. Its domain is eight months old, it has a handful of paying teams, and nobody owns SEO. It is a different fictional company from the Example Co. in other AAJ posts. Example Co., its pages and every detail about it are illustrative.

Step 1: Decide what to start now

Split SEO into its three jobs and the timing question mostly answers itself.

The technical basics. Make sure search engines can find, read and index your pages, and that each page says clearly what it is about. This is done once, takes an afternoon for a small site, and costs nothing. Do it now, whatever your stage.

Pages for people already looking for a fix. These are the searches of someone who already knows the problem and is comparing ways to solve it: "alternative to" a tool they use, your product type for a specific job, an integration they need. Few people make these searches, but the ones who do are close to buying, and your sales team can send the same pages to prospects. In my view, a startup should write the first few of these early, even before product-market fit, because they are useful twice.

A topic cluster. This is a set of linked pages that covers one subject in depth: one main page and several supporting posts. It is the slow, compounding part of SEO. I would start it only once your positioning has settled and you know the words your buyers use, because a cluster built on the wrong words is expensive to unbuild. If you are still changing who you sell to, finish the positioning work first.

How long until it works? No one can promise you a timeline, because it depends on how competitive your searches are, where your site starts and how much you publish. Google's own page on hiring an SEO doesn't give one either; it tells businesses to ask any SEO, "What kind of results do you expect to see, and in what timeframe?"^[2]. Its starter guide is equally careful: "Some changes might take effect in a few hours, others could take several months."^[3] Ask the question, and distrust a confident answer.

What you can control is what you judge, and when. In the first month, judge whether your pages are indexed. In the first few months, judge whether they are showing up in searches at all, and for which phrases. Only after that, judge clicks and sign-ups. Step 5 sets this out. This sequence is my judgement, not a benchmark.

Example Co., step 1 Fictional

The founders split the work in three. The technical basics go on this week's list. Three pages for active searchers go on this month's list: one comparing Example Co. with taking notes by hand, one for a CRM it connects to, and one for sales managers who review calls. The cluster waits until they have finished their customer interviews, because they are still deciding whether to sell to sales managers or to revenue operations teams.

Do this

Write down which of the three jobs you will start this month, and which you will deliberately not start yet.

Step 2: Fix the technical basics in an afternoon

Technical SEO has a reputation for being deep, and on a large site it is. On a startup's site of twenty or thirty pages, the basics fit in one afternoon, and they are the only part of SEO with no downside to doing early. Figure 2 is the checklist.

Set up Search Console and send it your sitemap. Search Console is Google's free tool for site owners, and it is where you will measure everything in Step 5. A sitemap is a file that lists your pages. Google suggests you "submit a sitemap in Search Console using the Sitemaps report", which also shows when Google read it and any errors, and it is honest about the limits: "Submitting a sitemap is merely a hint"^[4]. Links matter more for a new site. In Google's words, "the vast majority of the new pages Google finds every day are through links"^[3]. So link every new page from at least one page that is already indexed, such as your home page or blog index.

Check that your pages can be indexed. Use the URL Inspection tool in Search Console on your home page and your most important product page. Look for pages accidentally blocked from search, which often happens when a staging setting is carried over to the live site.

Give every page one clear title and description. Google builds the headline of your search result from sources that include the title. The short text under it usually comes from the page itself, and occasionally from the description^[3]. Write them for the person searching, not for a keyword tool: what the page is, for whom, in plain words. Two things you can skip: "Google Search doesn't use the keywords meta tag", and repeating a phrase many times is "against Google's spam policies"^[3].

Check speed, then keep it in proportion. Google's Web Vitals guidance sets good targets: the main content should appear within 2.5 seconds, the page should respond to a tap or click in 200 milliseconds or less, and the layout should barely shift (a CLS score of 0.1 or less), measured at the 75th percentile of page loads^[1]. Fix anything far off. Google does recommend good scores "for success with Search"^[5], but it also says good results in its Core Web Vitals report don't "guarantee that your pages will rank at the top of Google Search results", and that "trying to get a perfect score just for SEO reasons may not be the best use of your time"^[6]. So don't spend a sprint chasing a perfect score while your pages say little.

What can wait. Detailed schema markup beyond the basics, pages in several languages, and templates that generate hundreds of pages. One thing you can skip entirely for Google: its guide to AI search says you don't need "new machine readable files, AI text files, markup, or Markdown" such as llms.txt to appear in Google Search, "as Google Search itself doesn't use them"^[8]. Other AI engines are a separate question, covered in the GEO playbook.

Example Co., step 2 Fictional

The founders verify the site in Search Console and submit the sitemap. URL Inspection shows that the pricing page is blocked from indexing by a setting left over from the beta. They remove it, rewrite six page titles that all began with the company name, and link the new pages from the home page. The whole job takes one afternoon.

Do this

Verify your site in Search Console, submit your sitemap, and run the AAJ Crawl-to-Citation Scorer on your home page.

Free tool

Check one page against the basics

The free Crawl-to-Citation Scorer checks a page against the technical basics in Step 2 and the AI-answer basics, so you can see what to fix first.

Step 3: Pick keywords from your buyers' words

In my view, the most common mistake with AI products is aiming at the category name. If your category is new, few people search for it yet, because buyers do not know it exists. They search for the problem they have, for the way they do the job today, for alternatives to the tool they already use, and for things that fit their setup. Figure 4 sorts those into four kinds.

• Problem searches describe the pain in the buyer's words: "how to write a follow-up email after a sales call". They are early in the journey, so they suit guides and the cluster in Step 4.
• Old-way searches name the way the job is done today: "sales call notes template". The person has the problem and a workaround. A page that shows the old way and then a faster one fits them well.
• Alternative and comparison searches name a tool the buyer already knows: "[their current tool] alternative", "[tool A] vs [tool B]". They are close to buying. These are the pages for active searchers from Step 1.
• Fit searches check whether a product works in the buyer's setup: "[their CRM] call notes integration", "call notes for sales managers". Few people search them, and most who do are serious.

Where do the words come from? Not from a keyword tool first. Read your last sales-call notes, demo requests and support tickets, and write down the phrases buyers use before you explain anything to them. Once your pages are indexed, Search Console shows the queries people actually used to find you, including ones you never planned for. A keyword tool is useful after that, to compare phrases you already have. Treat its volume figures as estimates, and don't drop a phrase because the tool shows few searches: for fit and alternative searches, a handful of buyers a month can be worth more than a thousand readers of a general guide.

Two rules help a young domain. Pick phrases specific enough that you could write the best page on the web for them, which usually means longer phrases. And give each phrase one page, so you don't end up with three posts competing for the same search.

Example Co., step 3 Fictional

The founders reread notes from their last ten sales calls. Nobody said "AI sales notes", the term on their home page. Buyers said "I spend an hour after every call updating the CRM" and "our reps never log their calls". The founders build a list of 24 phrases, sort them into the four kinds, and find that only two use the category name.

Do this

List 20 phrases from your last five sales calls or demos, in the buyers' own words, and sort them into the four kinds in Figure 4.

Step 4: Build one cluster at a time

"How many blog posts do we need before SEO works?" is the wrong question, and the answer "as many as possible" is the wrong answer. The better question is: which one topic should we cover better than anyone, and what pages does that take?

Start with the pages for active searchers from Step 1, because they are few and close to revenue: a comparison with the old way, an "alternative to" page, a page per main use case, and a page per integration that matters. Each should be honest about who the product is not for. Buyers read these pages alongside the competitor's, and a fair page earns more trust than a one-sided one.

Then build one cluster, as in Figure 6: one main page that covers the topic as a whole, and five to eight supporting posts that each answer one narrower question, all linked to each other. The links do two jobs. They help readers move to the next question, and they show search engines that the pages belong together.

Why five to eight? It is my judgement, not a rule: enough to cover the questions a buyer asks about one topic, few enough that a small team can make every page the best answer it can give. Only start a second cluster when the first one is finished and linked.

Quality is the constraint, not count. Google describes the content it wants to reward as "created primarily for people, and not to manipulate search engine rankings"^[9], and its spam policies name the opposite: "Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users"^[10]. Fifty thin posts written in a month are a risk, not a strategy.

A public company's annual report puts it in the same plain terms. Freshworks' annual report for 2024 says its "digital and content marketing teams generate strong inbound demand through effective paid, social media, and SEO tactics that support website traffic growth and conversions", and that it relies "on efficient search marketing and word of mouth" so that users and small teams "discover, try, and purchase" its products^[11]. That is what the filing says, not an account of how Freshworks plans its content internally. The point for a startup: SEO sits next to paid and social as one way to bring in buyers, and it is judged on traffic that converts.

Example Co., step 4 Fictional

With the interviews done and sales managers chosen as the buyer, the founders pick one topic: "logging sales calls in the CRM". The main page explains the whole job. Six supporting posts answer one question each, such as what a sales manager should check in call notes, and every post links to the main page and to two others. They publish one post every two weeks and update the main page each time.

Do this

Draw one cluster on a single page: the main page, six supporting posts, and the links between them. Check that each post answers a phrase from your Step 3 list.

Step 5: Measure before the traffic arrives

It is easy to give up on SEO in the quiet months, when traffic is flat and nothing seems to happen. Something usually is happening, and Search Console shows it before your analytics do. Judge each stage on the signal that stage can produce, as in Figure 7.

• First, indexing. Are your new pages in Google's index? The page indexing report in Search Console tells you, and URL Inspection explains why a page is not.
• Then, impressions and queries. Are your pages showing up in searches, and for which phrases? The Performance report shows impressions before clicks come. The queries list is the most useful thing in it: phrases you planned for, and phrases you didn't, which go back into your Step 3 list.
• Then, clicks and sign-ups. Which pages bring visitors, and which of those visitors sign up or book a demo? This needs analytics set up properly; Tracking setup with GA4 and UTMs covers it.

The order is my judgement, not a benchmark, and I would not put months on it for your site. What I would do is look at the same three numbers once a month and write one line on what changed.

AI answers change what you measure, not the basics. Google says "The best practices for SEO continue to be relevant because our generative AI features on Google Search are rooted in our core Search ranking and quality systems", and for measurement it points to its own tool: "To measure how your content is performing in generative AI features on Google Search and Discover, use the Generative AI performance report in Search Console"^[8]. Some searchers will get their answer without clicking at all; Zero-click marketing covers how to plan for that, and the GEO and AEO Playbook covers being cited by other AI engines.

AI-written content is a tool, not a shortcut. Google says "Generative AI can be particularly useful when researching a topic, and to add structure to original content", and warns that "using generative AI tools or other similar tools to generate many pages without adding value for users may violate Google's spam policy on scaled content abuse"^[12]. In my view the test is simple: does the page contain something only your team knows, such as what you hear on sales calls, your product's real limits, or a worked example? If not, it adds nothing a buyer cannot get elsewhere. Why AI content doesn't sound like you and the AI Tool Policy cover how to use AI without losing that.

Example Co., step 5 Fictional

In the first month, the founders check that all new pages are indexed; two are not, because they were never linked. Over the next months the cluster starts to collect impressions, including for a phrase they had not planned: "call notes for new sales reps". They write a supporting post for it. Sign-ups from search are still few, and they expected that; their monthly note says what moved and what they will do next.

Do this

Set a 30-minute review on the first Monday of each month, with three numbers: pages indexed, impressions, and sign-ups from organic search. Write one line on what changed.

What changes by stage?

• Pre-seed — The technical basics and two or three pages for active searchers, written by the founder. No cluster yet; your positioning is still moving.
• Seed — A first cluster on one topic, built from the words in your sales calls, plus a monthly review of Search Console. One person owns SEO, even if it is a few hours a week.
• Series A and B — A second and third cluster, a refresh of the oldest pages every quarter, and a budget for a writer who sits in on sales calls. This is also the stage where programmatic pages can make sense; the AAJ programmatic SEO skill helps you check whether the build pays back first.

What mistakes should I avoid?

• Aiming at the category name of a new category nobody searches for yet.
• Waiting for traffic before setting up Search Console.
• Publishing many thin posts, written by AI or not, instead of a few good ones.
• Three pages competing for the same phrase.
• Judging SEO in week three, or on traffic alone in the first months.
• Choosing keywords from a tool before reading your own sales calls.
• Letting a staging setting block your pages from search.

Example Co. is a fictional company. Its pages, phrases and results are illustrative. This post is AAJ's judgement, not a guarantee of results.