---
title: "SEO Strategy: How I Build One in 7 Steps (AI Era) - Łukasz Rogala"
description: "An SEO strategy is a set of decisions, not a task list. Seven steps I use to build one, from measurement to AI Overviews and a plan. Data from 34 markets."
url: "https://rogala.pro/en/blog/seo-strategy/"
language: "en"
author: "Łukasz Rogala"
published: "2026-10-08T13:15:17+02:00"
modified: "2026-10-08T16:12:15+02:00"
---

[Blog](https://rogala.pro/en/blog/)

# SEO Strategy: How to Build One Step by Step

[Łukasz Rogala](https://rogala.pro/en/#about) · October 8, 2026 · 23 min read

**I build an SEO strategy in seven steps:**

1. I ask decision questions before I open the data.
2. I check the measurement before accepting the diagnosis.
3. I study demand and current Google results.
4. I check whether the click still comes back, that is, what AI Overviews and AI assistants do.
5. I write down hypotheses with a confidence level and a test.
6. I set priorities and a plan with an access list.
7. I oversee implementation and measure the results on a regular cadence.

**An SEO strategy is a handful of decisions: which demand you compete for, what you give up, in what order you act, and how you will know it works.** The task list (audit, keywords, content, links) only follows from those decisions. When a company has a task list without decisions, it has a work plan, but not a strategy.

**This post is itself an example of the method.** Before writing it, I checked demand for 2,427 keywords about SEO strategy and consulting in 34 markets in the 30 languages my site covers (6,088 market-keyword pairs with volume), and I pulled Google results for 86 queries, with AI Overviews and the “People also ask” questions. Data: DataForSEO, October 7, 2026. Pulling the data cost about USD 2.40, and the conclusions from the analysis are in the text.

## What is an SEO strategy?

**An SEO strategy is a diagnosis of what limits a site's visibility, a choice of how you will overcome it, and the coherent actions that follow.** This split comes from management, not SEO: in the book “Good Strategy Bad Strategy”, Richard Rumelt describes the kernel of every good strategy as a diagnosis of the main problem, a guiding policy and coherent actions. Kevin Indig, an SEO advisor to large technology companies, carried this model straight over to SEO and defines strategy as a way to overcome key challenges using competitive advantages.

> “Strategy is the most misunderstood topic in the SEO world and likely the most impactful one.”
>
> Kevin Indig, Growth Memo, August 19, 2024

**In practice a strategy starts with the question “what is blocking us”, not “what should we do”.** Indig writes that defining the challenge is the most important part of a good SEO strategy, because direction and all later actions rest on it. The diagnosis also determines what the company gives up. That is a hard moment, because every channel and every keyword cluster has someone on the team defending it.

**SEO strategy comes down to three decisions**

Three cards linked by arrows. 1. Diagnosis: what is really limiting search traffic growth. 2. Guiding policy: what we focus on and what we deliberately give up. 3. Coherent actions: what we do, in what order, with what access, and how we will know it worked. Below the cards is an example: the first two from one of my projects, the third as an illustration of a plan.

*Strategy, as Richard Rumelt describes it and Kevin Indig applies it to SEO, has three parts: diagnosis, guiding policy and coherent actions. The first two examples come from one of my projects; the third shows how such a decision turns into a plan.*

## How is an SEO strategy different from an audit and an action plan?

**An audit is the input to a strategy, a plan is its output, and tactics are the way of executing it.** Lists such as “10 best SEO strategies” actually describe tactics: link acquisition, content, page speed. Tactics without a diagnosis are shots in the dark, even if each one is correct.

| Concept | What question it answers | Example |
|---|---|---|
| Strategy | What limits us and what we give up | The cluster without an AI Overview first, the largest one postponed |
| Plan, schedule | What, who and when | Three posts in November, a template fix in December |
| Tactic | How to carry out a single action | Heading structure, internal linking, structured data |
| Audit | What state the site is in today | A list of technical and content errors with priorities |

## How to build an SEO strategy step by step?

**I build a strategy in seven steps: decision questions, measurement, demand and Google results, checking whether the click comes back, hypotheses, priorities with a plan, and implementation with measurement.** Between them are three decision gates, and every step ends with a file you can check. The gates exist so that I do not spend the client's time collecting data before it is clear why, and so that I do not hand over a document that nobody but the author has checked.

**How I build an SEO strategy: seven steps and three gates**

A timeline with seven steps: decision questions, baseline measurement, demand and live Google results, AI and clicks, hypotheses and falsification, priorities and plan, implementation and measurement. Between the steps are three decision gates: approval of plan and cost, thesis and scope, handover after cross-checking. Measurement feeds back into the hypotheses.

*Every step ends with a file you can check: a plan, a register of numbers, a keyword map, a ranking of hypotheses, a plan with an access list.*

### 1. Decision questions before I open the data

**Before I open any tool, I collect answers to one set of questions.** Who will make a decision based on the result? What counts as a conversion and how much is it worth? Which markets earn money? What changed on the site in recent months (a migration, a new template, a consent banner, a change of agency)? What do I definitely not promise, for example a return to the results from before a drop? The answers become a plan: the question, what is already known, what data I will check it with and how much collecting that data will cost. The first gate is approval of this plan.

### 2. Check the measurement before accepting the diagnosis

**Some “SEO problems” sit in the measurement, not in Google.** That is why I first check where the number comes from: which tool, since when, whether the analytics code or consent banner changed in that time, and whether the domain or redirects changed. I write down known breaks in comparability explicitly.

**Comparisons of Search Console data from 2025-2026 need special care.** Around September 10, 2025, the num=100 parameter, which showed 100 results on one page, stopped working. Google only commented that it is not a parameter it formally supports (Search Engine Roundtable), and in the same week many sites saw a drop in desktop impressions in Search Console (Search Engine Journal). The link between the two is an industry observation, not Google's confirmation. Separately, Google confirmed an impression-counting error from May 13, 2025 to April 27, 2026, which also affected CTR and average position, but not clicks (Google). That is why I do not compare impressions, CTR and position from this period year over year without a caveat, and for clicks I check only changes on the site itself.

**The second test at the start is whether bots can get into the site at all.** I check the server response for search engine, tool and AI assistant bots with a series of requests, because a firewall can block them only once the number of requests passes a threshold and a single test will show nothing. If a bot cannot get in, all further analysis measures nothing. I also check statements like “something is missing from the site” in the admin tools (Search Console, CMS), not only on the site from outside.

### 3. Demand and live Google results

**I group keywords into clusters by how Google arranges them, not by how they sound.** First I filter out keywords that do not fit the offer: every keyword in the plan has a named service or product it leads to and a page that answers it. Then I pull current Google results for every keyword, with the AI Overview text and the sources it cites, and I merge keywords that share URLs in the top 10. If two keywords share several URLs, one page is usually enough. If they share none, it is a strong signal that two pages are needed, even when they mean almost the same in English. I treat shared URLs as a hint, not a rule, and I always look at the result type too, because the composition of the top 10 changes between pulls.

**In the same step I run a competitor analysis.** I do not look at a list of rivals by name, but at the type of pages and domains that win in each cluster today, and at their link profile. That decides whether a cluster can be won with content, or whether the link gap has to be closed first.

**One page or two? Two queries, two different sets of results**

Two columns of Google results (up to ten in each). Each result is labeled with a page type: guide, service offer, directory or profile, other. Arrows would link URLs present in both sets.

*Google results in the United States, desktop, October 7, 2026 (DataForSEO), result type assessed manually. “seo consultant” leads mostly to directories and profiles, while “seo strategy” leads only to guides.*

**This is how I decided to split this material into two posts.** In the USA, for “seo strategy” guides win (9 of 9 organic results in the top 10), while for “seo consultant” the results are mostly directories and profiles (5 of 9), with only 2 guides and 1 service offer. Shared URLs in the top 10: zero. In the 19 markets where I checked the same pair of keywords (including the USA), the result was the same: not a single shared URL, guides in 151 of 165 organic results for strategy and in only 55 of 180 results for a consultant. Each keyword is one pull from October 7, 2026, so the strength of this conclusion comes from repeating it in 19 markets, not from a single comparison. That is why this post answers questions about strategy, and I write separately about when a consultation is needed: [When Do You Need an SEO Consultant?](https://rogala.pro/en/blog/when-do-you-need-an-seo-consultant/)

**Every keyword in the analysis must have a source: tool volume, Search Console impressions or presence in the results.** A keyword added “off the top of your head” is an assumption, not data. In one of my tests (August 2026), 9 of 14 service names that I derived by hand from page URLs had no demand at all, and a conclusion built on them did not survive checking.

### 4. Does the click still come back

**Before the content plan, every cluster gets one of five categories.** I combine the presence of an AI Overview and the list of cited sources with real demand from Search Console.

**Five situations in a keyword cluster when AI is in the results**

A table of five situations: no AI Overview, AI answers and cites us, AI answers and does not cite us, closed by official sources, out of reach. For each one: what it means and what we do.

*I apply this split to every cluster before the content plan, so the order changes before anyone starts writing.*

**This split changes the order of the plan before anyone starts writing.** A cluster without an AI Overview and a cluster where AI answers and does not cite us go first, because there the work turns into clicks or citations faster. In the second of these groups the page usually already ranks and lacks a citation, so the fix is cheaper than new content. When AI Overviews are held by government sources or big sites, I postpone the head term even if it has the largest volume, and I test its variants separately.

**I build the question set for measurement in AI assistants from two bands.** The growth band is keywords at positions 11-30, where the site should climb higher. The defense band is top 10 keywords that already bring traffic. In one project (September 2026, Search Console data), a set made up only of the 11-30 band would have missed, in one keyword family, about 84% of impressions from positions 1-30 (about 19,000 of 23,000), and 4 of 7 brand citations in the AI Overview on the day of measurement came from the top 10. Such a set would measure gains and not notice losses.

**Separately I check whether the model searches the web at all.** In my test through the API (October 4, 2026), Gemini 3.8 Flash did not use search in 5 of 5 how-to questions, Gemini 3.1 Pro in 3 of 4, and the older Gemini 2.5 Flash searched on every question. These are a few questions, so the result is indicative. Where a model answers from memory, position in Google changes nothing. What counts is what was on the web about the brand before the model's training cutoff, which is different work and a different horizon, so I report both layers separately.

### 5. Hypotheses with confidence levels

**Instead of one cause I give several hypotheses, each with a confidence level and a test that could refute it.** I use three levels: “measured” (client data or a test), “probable” (public data consistent across several sources) and “to check” (a hypothesis with a test). Without a measurement I do not write “it dropped because of a Google update”, but “it coincides with an update”, and what would settle it.

> “That does not mean experience is useless. It means experience should produce hypotheses, not certainty.”
>
> Will Critchlow, founder of SearchPilot, SearchPilot blog, August 28, 2026

**A hypothesis has a test that can refute it**

Three hypothesis cards from my projects. Each has a test result (refuted or confirmed), a test that could have refuted it, and what the result changed in the plan.

*Examples from three projects, data generalized. This is how I record every causal claim before it goes into a client document.*

**I record every number in a register with a denominator, a source and a date, and every claim in a claims register with a test.** Before a document is sent, a script compares the text with the register and stops it if they differ. Comparative claims (“competitors did not drop, we did”) I check on the whole group, not on a few examples.

**The same rule applies to my own tests.** When I compare two signals, I measure both under the same conditions and on a sample split that can be reproduced. Agreement of calculations with the register is not enough if the way the data was collected favors one of the signals, so I recompute the result from scratch from raw data before launching anything based on it.

### 6. Priorities and plan

**The analysis produces recommended actions, not observations like “something is rising, something is falling”.** Each recommendation specifies where to make the change (a URL or a template), the content of the change, a check in current Google results that the change has a chance, and the gain as a range, not a single number. A single number from a model implies a level of certainty that the model does not have.

**First I take on what almost works already.** Keywords at positions 11-20 with high demand, pages with many impressions and low CTR (the share of impressions that end in a click), and “AI answers and does not cite us” clusters. New content comes after them, because a new page rarely reaches the top 10 fast: in an Ahrefs study, 1.74% of a million new URLs got there within a year (September 2023), and 6.11% in a separate sample of 2 million URLs with English content (Ahrefs).

**A plan always has an access list and stages.** Each access requirement (Search Console, analytics, CMS access, server logs) has an owner, a deadline and a plan in case it is not available. I split work into stages with an hour cap, and the client can stop after each stage.

**The second gate is a thesis with a confidence level and the scope of work, and the third is cross-checking.** Several independent reviewers read the document: one recomputes the numbers from raw files, another checks the graphics, a third tries to refute the claims. After the fixes I run a second round only on the difference, because fixes introduce errors too. In one project the second round found errors introduced only by the fixes, for example a wrong denominator in a client guide (79 instead of 76 links), which my own check script did not catch. Two rounds reduce the risk of a mistake but do not remove it, and that is what I tell the client.

### 7. Implementation and measurement on a regular cadence

**Measurement has a rhythm and ends with a task list for the next period, not just charts.** Every week I check Search Console data once the previous week is complete. Every month I set directions for the next month, and every quarter I compare the trend with the work done, including where it did not work. Every tested hypothesis has a metric, a baseline, a failure criterion and a measurement date in the register. A refuted hypothesis is also a result: for example, I checked whether more H2-H4 headings increase the chance of top 3, and on a small sample (Google results for the keywords of two companies, in English, September 2026) they did not.

**I do not measure success by traffic alone.** Rand Fishkin writes it plainly: “Traffic can fall precipitously even as revenue rises” (SparkToro). Will Critchlow warns against treating the year-over-year traffic trend as the measure of success in e-commerce (SearchPilot). That is why a strategy records from the start what counts as a conversion, and the report shows traffic next to customer inquiries, sales and AI visibility.

**Want me to go through these steps with your site?** Tell me your goal and what you think blocks growth. I will reply with the step I would start from and what I will need.

[Get in touch](https://rogala.pro/en/#contact) · [LinkedIn](https://www.linkedin.com/in/lukaszrogala/?locale=en_US) · [WhatsApp](https://wa.me/48570081225)

## What should an SEO strategy document contain?

**An SEO strategy document should answer seven questions: where we are going, what blocks us, what we give up, which pages will serve which keywords, what we do and in what order, what we need for it, and how we will know it worked.** If one of these elements is missing from the document, it is usually a sign that it is an audit or a schedule, not a strategy.

| Element | What question it answers | Example |
|---|---|---|
| Goal and measure | What result we change, from what level and by when | Customer inquiries from search: today's level, target and deadline |
| Diagnosis | What limits visibility today | Service pages lose to guides on keywords that sell |
| Guiding policy | What we give up and why | We postpone the cluster whose AI Overview is held by government sources |
| Keyword and page map | Which page answers which keywords and how pages link to each other | A keyword cluster assigned to one page, with links from guides |
| Actions and order | What we do first and what later | Fixing pages at positions 11-20 before new content |
| Resources, budget and access | Who does what, how much it costs and what access is needed | Developer and author hours per stage, a cost cap, access to Search Console and the CMS |
| Measurement and rhythm | When and how we check the effect | A monthly review, a cluster compared with a cluster without changes |

## What does AI change in SEO strategy?

**AI Overviews and AI assistants did not invalidate SEO, but they changed its arithmetic: a high ranking no longer guarantees a click.** In my analysis an AI Overview appeared in 69 of 86 checked Google results, and for the SEO strategy keyword in all 19 markets where I checked it. Independent measurement of user behavior comes from Pew Research Center, which tracked the browsers of 900 American adults for a month.

- 8% vs 15%The share of visits to a results page that ended with a click on a regular result when Google showed an AI summary, versus visits without a summary. 68,879 searches, USA, March 2025.Pew Research Center
- 1%The share of visits with an AI summary that ended with a click on a link cited in the summary itself.Pew Research Center
- -19.9%The median change in clicks across 1,442 Polish domains with stable positions, August-October 2025 versus 2024. This is the change in the domains' whole traffic, not a result only for keywords with an AI Overview. Separately, Senuto reports that an AI Overview appeared then for 28.95% of keywords in its database.Senuto

**The industry disagrees about the scale of this phenomenon, and it is worth knowing both sides of the dispute.** Google claims that total traffic from its search engine to sites is stable. Researchers and tools measure drops in clicks for keywords where an AI Overview appears.

> “Overall, total organic click volume from Google Search to websites has been relatively stable year-over-year.”
>
> Liz Reid, head of Google Search, Google blog, August 6, 2025

**Both sides can be right at once, because they measure different things.** Google talks about the sum of clicks from all queries and gives no numbers in that post, and a stable sum with a growing number of queries means fewer clicks per query. The studies talk about clicks for specific queries with an AI Overview. The numbers on the CTR drop also differ by method. Ahrefs reports that with an AI Overview the CTR of the first result is 58% lower than the trend for keywords without an AI Overview would suggest (300,000 keywords, December 2025 versus December 2023, Ahrefs). Sistrix, in an analysis for Germany from February 2026, compared the CTR of the first position in regular results (about 27%) with the CTR with an AI Overview (about 11%), and estimated the total loss of clicks across all keywords at 6.6%, with an AI Overview on a little over 20% of keywords (Sistrix).

**Clicks also fall where there is no AI Overview.** According to Seer Interactive, the CTR of keywords without an AI Overview fell from 2.72% in June 2024 to 1.62% in September 2025 (groups set by the presence of an AI Overview in September 2025), so the whole drop cannot be attributed to the AI Overview alone (Seer Interactive). That is why in a strategy I do not adopt one number for the whole site, but measure each keyword cluster separately, as in step 4, and compare it with clusters without an AI Overview, not only with its own past. Rand Fishkin thinks better SEO alone will not reverse the general decline in clicks, but he points out where SEO still pays off: branded queries, local businesses and queries with clear purchase intent (SparkToro). That is why I assess profitability for each group of queries separately, and I start the plan with those where the click still comes back.

**In assistants such as ChatGPT, Gemini or Perplexity I measure something other than position.** The same site gets a different result when someone asks the assistant a question in a sentence and when they type a bare keyword, so I measure both. On top of that, assistants' answers are unstable: in a SparkToro experiment from November and December 2025 (2,961 answers, 600 volunteers), the chance that two answers give the same list of brands was, for two of the three tools studied, less than 1 in 100 (SparkToro). That is why I always report AI visibility as a share of my set of questions and keywords, with the date and the model version, and not as a “share of the AI market”.

**Whether optimizing for AI is a new discipline or still SEO is also disputed.** Google writes that, from its perspective, “optimizing for generative AI search is optimizing for the search experience, and thus still SEO” (Google Search Central). Mike King of iPullRank puts it differently: the only part that is “just SEO” is the fact that it is being done incorrectly (iPullRank). My position is practical: the foundation is the same (the site must be accessible to bots, understandable and better than the alternatives), but measurement and priorities are different, so a strategy must account for them from day one.

## How to plan an SEO strategy across several markets?

**Start the plan for each market with its own data, because a decision made in one market does not carry over to others by itself.** In my keyword set for this post, in the 15 largest non-English markets, keywords about SEO strategy in the market language are only 15% of demand, and 85% is SEO consultant, consulting and advisory. In the USA strategy is 31%, in India (English keywords) 71%, in Italy 5%, in Turkey 2%. The same topic has a different shape in every market.

**What people search for in 18 countries: SEO strategy or a consultant**

Horizontal bars by country. Each bar splits monthly demand for keywords in the market language into two parts: SEO strategy, and SEO consultant, consulting and advisory. Next to it, total searches per month.

*Average monthly volume of keywords in the market language, DataForSEO Labs, October 7, 2026. In non-English markets English keywords are excluded; in the USA, the United Kingdom and India English keywords are counted. Without duplicate volumes merged by Google Ads. Shown: the 18 markets with the most demand in the set (each at least 100 searches a month), without Canada, Australia and Hindi keywords. The result depends on the choice of keywords.*

**The intent of very similar keywords in one language also differs.** In Poland, for “konsultacje seo” company offers win, and for “konsultant seo” profiles of specific people in company directories and social networks. Such different result types are a signal that one page will struggle to serve both intents, and a page translated from another market may answer neither.

**I check service names and Google results for each market separately.** I check service names in the market language before the URLs are built, and current Google results before I carry content or a redirect over from another language version. Examples of how much names of the same service can differ are in the post on [when you need an SEO consultant](https://rogala.pro/en/blog/when-do-you-need-an-seo-consultant/).

### This article in other languages (26)

- [PolskiStrategia SEO: jak ją zbudować krok po kroku](https://rogala.pro/blog/strategia-seo/)
- [DeutschSEO-Strategie entwickeln: So gehen Sie Schritt für Schritt vor](https://rogala.pro/de/blog/seo-strategie/)
- [EspañolEstrategia SEO: cómo crearla paso a paso](https://rogala.pro/es/blog/estrategia-seo/)
- [FrançaisStratégie SEO : comment la construire étape par étape](https://rogala.pro/fr/blog/strategie-seo-comment-la-construire/)
- [ItalianoStrategia SEO: come costruirla passo dopo passo](https://rogala.pro/it/blog/strategia-seo/)
- [PortuguêsEstratégia de SEO: como montar passo a passo](https://rogala.pro/pt/blog/estrategia-de-seo/)
- [NederlandsSEO-strategie: zo bouw je die stap voor stap](https://rogala.pro/nl/blog/seo-strategie/)
- [ČeštinaSEO strategie: jak ji vytvořit krok za krokem](https://rogala.pro/cs/blog/seo-strategie-krok-za-krokem/)
- [DanskSEO-strategi: sådan bygger du den trin for trin](https://rogala.pro/da/blog/seo-strategi/)
- [SvenskaSEO-strategi: så bygger du den steg för steg](https://rogala.pro/sv/blog/seo-strategi/)
- [Norsk bokmålSEO-strategi: slik bygger du den steg for steg](https://rogala.pro/nb/blog/seo-strategi/)
- [SuomiSEO-strategia: näin rakennat sen vaihe vaiheelta](https://rogala.pro/fi/blog/seo-strategia/)
- [RomânăStrategie SEO: cum o construiești pas cu pas](https://rogala.pro/ro/blog/strategie-seo/)
- [MagyarSEO stratégia: hogyan építse fel lépésről lépésre?](https://rogala.pro/hu/blog/seo-strategia/)
- [ΕλληνικάΣτρατηγική SEO: πώς να τη δημιουργήσετε βήμα προς βήμα](https://rogala.pro/el/blog/stratigiki-seo/)
- [УкраїнськаSEO-стратегія: як побудувати її крок за кроком](https://rogala.pro/uk/blog/seo-stratehiia/)
- [العربيةاستراتيجية سيو: كيف تبنيها خطوة بخطوة](https://rogala.pro/ar/blog/istiratijiyya-seo-khatwa-bikhatwa/)
- [עבריתאסטרטגיית SEO: איך בונים אותה שלב אחר שלב](https://rogala.pro/he/blog/astrategiyat-seo-shlav-ahar-shlav/)
- [TürkçeSEO Stratejisi: Adım Adım Nasıl Oluşturulur?](https://rogala.pro/tr/blog/seo-stratejisi-nasil-olusturulur/)
- [日本語SEO戦略の立て方：7つのステップで進める方法](https://rogala.pro/ja/blog/seo-senryaku-tatekata/)
- [한국어SEO 전략: 단계별로 만드는 방법](https://rogala.pro/ko/blog/seo-jeollyak-mandeuneun-bangbeop/)
- [繁體中文SEO 策略怎麼做？一步一步建立](https://rogala.pro/zh-hant/blog/seo-celue-zenme-zuo/)
- [简体中文SEO 策略怎么做？一步一步制定](https://rogala.pro/zh-hans/blog/seo-celue-zenme-zuo/)
- [ไทยกลยุทธ์ SEO: วิธีสร้างทีละขั้นตอน](https://rogala.pro/th/blog/konlayut-seo-khan-ton/)
- [Tiếng ViệtChiến lược SEO: cách xây dựng từng bước](https://rogala.pro/vi/blog/chien-luoc-seo/)
- [Bahasa IndonesiaStrategi SEO: Cara Menyusunnya Langkah demi Langkah](https://rogala.pro/id/blog/strategi-seo-langkah-demi-langkah/)

## What mistakes to avoid in an SEO strategy?

1. **A forecast from someone else's result.** Measures such as Traffic Potential in Ahrefs show the traffic of the page that ranks first today, across all its keywords. It is an estimate of someone else's page traffic, not a forecast for yours.
2. **A percentage without a denominator.** “+62%” without a base number is not evidence. I give a share with its denominator (“59 of 537”), and a change with its base and period.
3. **One cause of a drop.** “It dropped because of a Google update” without checking whether the same jump appears in a second source and on sites from the same niche.
4. **A plan built only from third-party tools.** Google reminds us that third-party tools do not have access to its internal ranking data (Google Search Central). Search Console and analytics come before estimates.
5. **A “growth on growth” forecast.** Kevin Indig notes that using last year's result as the starting point of a forecast assumes that everything that worked will work again (Growth Memo). I give forecasts as ranges with three scenarios.
6. **Volume from an annual average in season.** A seasonal keyword can average 7,800 searches a month over a year and about 26,500 in August 2026, during the season. The annual average is used to value potential, and monthly volume to plan publication before the season.
7. **`noindex` on pagination.** Pagination pages are sometimes the bot's only path to deep subpages. I leave them as `index, follow` and shorten the pagination or add other link paths.
8. **The same decision in all markets.** Intent and service names differ even between neighboring countries.

## When should you hire a consultant for your SEO strategy?

**When the decision is hard to reverse, or the company has no time to analyze data.** Typical cases are a migration, entering a new market, a traffic drop with no obvious cause, and a situation where the team works on the site but nobody sets priorities. I describe all of them in a separate post, together with the questions worth asking a consultant before you sign a contract: [When Do You Need an SEO Consultant?](https://rogala.pro/en/blog/when-do-you-need-an-seo-consultant/)

**Wondering whether your SEO strategy makes sense?** Tell me what you already do and what results you have. I will reply whether a consultation is enough or a full strategy is needed.

[Get in touch](https://rogala.pro/en/#contact) · [LinkedIn](https://www.linkedin.com/in/lukaszrogala/?locale=en_US) · [WhatsApp](https://wa.me/48570081225)

## Frequently asked questions about SEO strategy

### Is SEO dead now with AI?

**No, but the arithmetic is different.** An AI Overview reduces the number of clicks for some queries (in the Pew study, 8% versus 15% of visits to a results page ended with a click on a regular result), and for others it does not appear at all. AI assistants also use search results when they search the web. A strategy that measures each cluster separately usually finds keywords where the click still comes back, and keywords where the goal is a brand citation, not a visit.

### Can I do an SEO strategy myself?

**Yes, go through the seven steps in this post in a shortened version.** The minimum is Search Console and analytics data, a manual check of current Google results for the main keywords (who wins, whether there is an AI Overview) and writing down what you give up. A common mistake in a do-it-yourself strategy is a list of tactics without a diagnosis.

### What tools do you need for an SEO strategy?

**Search Console and analytics are the foundation, and for demand research you need one keyword tool.** Search Console shows which keywords the site already appears for, analytics (for example GA4) shows conversions, and a keyword tool (Ahrefs, Semrush or Google Ads Keyword Planner) shows demand. For technical checks, a program that crawls the site the way a bot does is useful, for example Screaming Frog. Third-party tools give estimates, so Search Console data comes before them.

### What are examples of SEO strategies?

**A strategy is a choice, not a tactic, and my examples are choices about order.** In one project the plan started with the cluster that had no AI Overview, and the largest cluster, whose AI Overview was held by government sites, was postponed. I also prioritize improvements to pages at positions 11-20 before writing new content. Lists of “best strategies” usually describe tactics such as links, content or page speed.

### What is the 80/20 rule in SEO?

**In my practice it is an order: first what almost works already, then new content.** What almost works is keywords at positions 11-20, pages with high impressions and low CTR, and content that ranks but has no citation in the AI Overview. A small share of changes usually delivers most of the effect, but which share, only the data of a specific site shows.

### How much does SEO cost per month?

**According to an Ahrefs survey of 439 SEO service providers, a monthly retainer usually costs between USD 250 and USD 10,000.** Ahrefs also estimates from providers' price lists that 63% of companies pay between USD 500 and USD 5,000 (article updated in August 2024). This is not a study of client budgets or of your market. The spread comes from scope: a strategy with implementation checks costs one amount, and full execution with content and links another. I write more about consulting costs in the post on [when you need an SEO consultant](https://rogala.pro/en/blog/when-do-you-need-an-seo-consultant/).

### How long does it take to see results from an SEO strategy?

**Google says some changes take effect in a few hours and others in a few months.** Technical fixes and improvements to pages that already rank usually show first. New pages need more time: in an Ahrefs study, from 1.74% to 6.11% of new URLs reached the top 10 within a year, depending on the sample. That is why a plan has milestones every month, not a single date for evaluating results.

### How often should you update an SEO strategy?

**I review the diagnosis and directions every quarter and the plan for the next month every month.** Sooner when something big changes: Google results for key clusters, the company's offer, the site (a migration, a new template) or an AI Overview appears where there was none.

## Where to start?

**If you have a site and want to know whether your SEO strategy makes sense, answer three questions.** What blocks growth (diagnosis), what you give up (guiding policy) and how you will know it works (measurement). If you cannot answer one of them from data, it is a good topic for an [SEO consultation](https://rogala.pro/en/#contact) or an SEO audit.

External sources (checked October 7, 2026):

- Growth Memo, Kevin Indig: How to craft a winning SEO strategy (growth-memo.com)
- Growth Memo: The baseline problem in SEO forecasting (growth-memo.com)
- Google Search Central: Do you need an SEO? (developers.google.com)
- Google Search Central: Optimizing for generative AI search (developers.google.com)
- Google Search Central: Evaluating SEO advice and third-party services (developers.google.com)
- Google Search Central: SEO Starter Guide (developers.google.com)
- Google: Data anomalies in Search Console (support.google.com)
- Google: AI in Search is driving more queries and higher quality clicks (blog.google)
- Search Engine Journal: Google modifies search results parameter (searchenginejournal.com)
- Search Engine Roundtable: Google comments on Search Console impressions dip (seroundtable.com)
- Pew Research Center: Google users are less likely to click on links when an AI summary appears (pewresearch.org)
- Senuto: AI Overviews w Polsce 2.0 (insight.senuto.com)
- Ahrefs: AI Overviews Reduce Clicks by 58% (ahrefs.com)
- Sistrix: AI Overviews in Deutschland (sistrix.de)
- Seer Interactive: AIO impact on Google CTR, September 2025 update (seerinteractive.com)
- SparkToro: AIs are highly inconsistent when recommending brands (sparktoro.com)
- SparkToro: In 2026, less than one third of Google searches still send a click (sparktoro.com)
- iPullRank: How AI Mode works (ipullrank.com)
- SearchPilot: When SEO best practices backfire (searchpilot.com)
- SearchPilot: When is a click not a click? (searchpilot.com)
- Ahrefs: SEO pricing (ahrefs.com)
- Ahrefs: How long does it take to rank (ahrefs.com)

My own data: DataForSEO Labs and live Google results, 34 markets, 86 queries, October 7, 2026.

[Łukasz Rogala](https://rogala.pro/en/#about)

SEO strategist and consultant

Łukasz Rogala is an SEO strategist and consultant, CEO of VerticalSEO and co-founder of Linko. Since 2014 he has delivered 360° SEO: strategy and audits, technical SEO, content and link building, semantic SEO and brand visibility in AI search (GEO).

[About](https://rogala.pro/en/#about) · [Talks](https://rogala.pro/en/#talks)

Read next

[When Do You Need an SEO Consultant?](https://rogala.pro/en/blog/when-do-you-need-an-seo-consultant/)