Check this page with an assistantOpens a chat asking it to summarise this article and name the evidence behind each claim.
Claude opens with the prompt on your clipboard: Anthropic does not support prefilled prompts on the web, and we would rather copy it than ship a button that drops it.
Three Requirements, Not One Discipline
A search system has to find your page, understand what it is about, and decide it is a good answer for a specific question. Those are three separate requirements. They fail independently, they look identical from the outside, and they have completely different fixes.
That framing does more work than any definition. When someone says their SEO is not working, the useful question is which of the three is failing. A page nobody can find and a page that is found and outranked are both invisible in a report and share nothing else. The full sequence, with the coverage states that tell you which stage you are at, is in how Google Search actually works.
It also explains why generic advice disappoints. A checklist written for someone stuck at understanding is useless to someone stuck at discovery, and most published advice does not say which problem it solves. Diagnosing first is not a preliminary step; it is most of the work.
What the Work Actually Consists Of
Four activities cover it. Making the site reachable and readable. Deciding what to publish based on what people search. Making individual pages the best available answer. And earning the corroboration that makes you credible to systems and people.
| Activity | What it answers | Where to start |
|---|---|---|
| Technical foundation | Can systems reach and read the pages? | Indexing diagnostics |
| Demand research | What do people actually search for? | Keyword research and search intent |
| Page quality | Is this the best answer available? | Service pages and comparison pages |
| Corroboration | Does anyone else vouch for you? | Links and citations |
| Measurement | Is any of it working? | Search Console |
Notice what is absent. There is no activity called optimising keyword density, no step where you submit your site to search engines, and nothing about meta keywords. Those belong to a version of search that ended years ago, and the ones still circulating are collected in SEO beliefs that were true once.
What Changed, and What Did Not
Search now spans several surfaces: ordinary results, generated answers, assistants, and the discovery that happens inside marketplaces and stores. What did not change is the underlying requirement, because Google's AI features are served from the same index as everything else.
That single fact settles a lot of anxious strategy. A page that is not indexed cannot be quoted in a generated answer, because there is nothing to quote from. No special file, markup, or optimisation substitutes for being in the index, which is why the tactics being sold as AI-specific mostly are not, as examined in our review of llms.txt.
What genuinely changed is the shape of the reward. Some queries resolve without anyone clicking, which affects what winning a query is worth rather than how you win it. Some visitors now arrive already informed by an assistant, which changes what the landing page should say. And which sources get cited is disproportionately third-party rather than brand-owned. Those are covered in zero-click searches, converting AI traffic, and earning third-party citations.
The Order Things Have to Happen In
Sequence matters more than effort here, because later steps depend on earlier ones being true. Publishing more content on a site nobody can crawl is the most common way to spend a quarter and change nothing.
- Confirm you are actually indexed.
Inspect a few real URLs and read the coverage state. Everything downstream assumes this, and a surprising number of sites discover here that Google has never seen most of their pages. The launch sequence is in the first 90 days.
- Find out what people search, in their words.
Not the vocabulary your industry uses internally. Your own Search Console, your inbox, and autocomplete are better sources than any tool, per long-tail keywords.
- Match the page type to the query.
Read the results page for your target query and see what format wins. A guide competing where every result is a comparison page is not competing badly; it is in the wrong category entirely.
- Build the pages closest to a decision first.
Service, product, and comparison pages before educational content. They have less volume and produce the customers, which is the trade most content plans get backwards.
- Earn corroboration from elsewhere.
Links, reviews, listings, and mentions on sources that get cited. This is slower than everything above and it is what separates sites that plateau from ones that compound.
- Measure, then keep it stable.
One method, applied consistently, so a change means something. The framing is in the reporting framework.
What SEO Cannot Do
It cannot create demand that does not exist, cannot make a page deserve to rank that does not, and cannot deliver on a schedule. Being clear about the limits is what separates a realistic plan from the ones that produce disappointment on a predictable timeline.
The demand limit is the one most often missed. If nobody searches for what you sell, search is the wrong channel and no amount of skill changes that. Checking whether the demand exists before building a programme around capturing it is a twenty-minute exercise that occasionally saves a year.
The timing limit is the one that damages relationships. Search cannot be switched on for a launch and cannot be relied on for a specific month. Where you need traffic on a date, that is what paid placement is for, and the honest comparison is in SEO vs Google Ads. The realistic timescales are in how long SEO takes.
Questions People Ask About SEO
- What is SEO in simple terms?
Making it possible for search systems to find your pages, understand what they are about, and choose them as a good answer for someone's question. Those are three separate requirements that fail separately, and most problems described as SEO problems are actually a failure at one specific one of them.
- Is SEO still worth doing with AI answers?
Yes, for a mechanical reason: AI features in Google are served from the same index as ordinary results, so a page that is not indexed cannot be quoted in a generated answer either. Everything upstream of the answer works the same way. What changed is that some queries now resolve without a click, which affects what winning is worth rather than how it works.
- How long does SEO take to work?
Discovery can happen in days. Indexing can follow quickly or not at all. Competitive ranking generally settles over months as evidence accumulates, and search becoming a meaningful revenue channel usually takes longer still. The question has no single answer because each stage runs on its own clock.
- Can I do SEO myself?
The first pass, yes, and you will understand your own situation better for having done it. Confirming pages are indexed, matching content to what people search, and completing your business profile are all things an owner can do in a weekend. Where outside help earns its cost is judgement about ambiguous trade-offs and implementation capacity you do not have.
- What is the difference between SEO and paid search?
Paid placement stops the moment you stop paying; organic visibility persists and compounds, but takes months to establish and cannot be switched on for a launch. They fail differently too: paid is capped by budget, organic is capped by whether you genuinely deserve to be the answer.
