What search engine optimization actually is
Search engine optimization is the practice of making a website into an answer a search engine is willing to give. That is a narrower job than it sounds and a more technical one than most descriptions of it admit. It involves persuading nobody. There is no relationship to cultivate at Google, no submission form that matters, and no lever that produces rankings on demand. There is a set of systems — a crawler, an indexer, a ranking stack, and now a generative layer sitting on top of all three — each of which behaves predictably when you give it something predictable to work with.
I have been practicing SEO since 1996, which means I have watched the field acquire and then quietly shed a great many theories. The durable part is small. A search engine has to reach a page, work out what it is about, decide it is a better response to a particular query than the alternatives, and be reasonably confident the source is trustworthy. Every tactic anyone has ever sold as SEO is an attempt to influence one of those four things.
The reason the discipline has an uneven reputation is that the influence is indirect and the feedback loop is slow. You change something, weeks pass, a number moves, and nobody can prove the two events are related. That gap between action and evidence is where bad advice lives, and it is why so much of good SEO work is really measurement discipline wearing a different hat.
Find, understand, choose
It helps to keep the three stages separate, because they fail separately and the fixes are not interchangeable.
Crawling
Crawling is a program requesting your URLs. It finds them by following links, reading sitemaps, and revisiting things it has seen before. It has a finite appetite, it obeys the instructions your server gives it, and it does not click on anything. A site that generates an unbounded number of URLs out of filters, sorting parameters and session identifiers will spend its crawler attention on combinations no human ever requested, while the pages that earn money get revisited monthly.
Indexing
Indexing is the decision to store a fetched page and make it eligible to appear. It is a decision, not an automatic consequence, and search engines decline constantly — because the page duplicates another page, because it is thin, because a canonical tag or a directive told them to, or because the page was judged not worth the storage. Blocking a URL in robots.txt prevents crawling, not indexing, which is the single most persistent misunderstanding in the field.
Ranking
Ranking is the ordering of the eligible set for a specific query, in a specific place, on a specific device, at a specific moment. It is not a property of your page. It is the output of a comparison, which means your position can change because a competitor changed, because the query's intent was reinterpreted, or because the results layout was redesigned and the thing you rank for is no longer where anyone looks.
What a search engine is actually ranking
Search engines rank URLs, not websites, and they rank them against an interpretation of what the person searching wants. Before optimizing anything for a term, look at what is already ranking for it. If the entire first page is product listings, a blog post will not win it. If it is all definitional explainers, a category page will not win it either. The existing results are the clearest available statement of what the engine believes the query means, and arguing with that statement is expensive.
Beyond intent, the signal groups have been stable for a long time:
- Content that matches the question — not keyword density, which stopped mattering long ago, but genuine coverage of what a person asking that question needs to know, on a page structured so the answer is findable.
- Links from other sites — still the strongest independent endorsement available, and still the signal most aggressively manipulated, which is why the systems that evaluate links have become far better at discounting than at penalizing.
- Entity clarity — whether the engine can tell who you are, what you sell, where you operate, and that the same organization is being referred to consistently across the web. This has grown more important as results moved toward answers.
- Technical access — the boring prerequisite. It does not win rankings. It loses them.
A useful test for any proposed tactic: does it change what the page genuinely is, or does it change only what the page claims to be? The first kind of change ages well. The second gets discounted, usually within a couple of algorithm cycles, and sometimes with a penalty attached.
What has changed since 1996, and what has not
The changes are real and they are worth naming, because people who learned SEO in one era keep applying it in the next. Directory submissions and meta keyword tags stopped mattering. Keyword-stuffed doorway pages stopped working and then started attracting manual penalties. Exact match domains lost most of their inherent advantage. Purchased links moved from a standard tactic to a liability. Mobile rendering became the primary way pages are evaluated rather than a secondary consideration. Encryption became an expectation. Page experience measurements arrived. And results pages filled up with features — local packs, product panels, video carousels, and now generated answers — that take clicks away from the blue links beneath them.
What has not changed is more useful to hold on to. A page a crawler cannot reach still cannot rank. Duplicate content still splits signals between URLs that should be one URL. An endorsement from a genuinely independent site is still worth more than a hundred from sites that endorse everyone. And anything cheap, scalable and available to everyone is on a timer, because the value of a signal to a search engine is inversely proportional to how easily it can be manufactured.
The arrival of generated answers has not overturned that logic so much as raised the stakes on the entity work. A system composing an answer has to decide which sources it trusts on the subject. That is the same question search has always been asking, with less room on the page for second place.
Measurement, and the limits of it
Most arguments about SEO are actually arguments about measurement, so it is worth being precise about what each source can tell you.
Search Console is the site's own record of impressions, clicks and average position. It is the most reliable thing available, with real edges: history ages out, average position is weighted by impressions rather than being a true rank, and queries below a privacy threshold are omitted entirely, which quietly undercounts the long tail.
Analytics tells you what happened after the click, and is affected by consent banners, ad blockers, bot filtering and tagging changes far more than most reports acknowledge. A drop in sessions with no corresponding drop in orders is a tracking story, not a search story.
Rank tracking tools take samples. They record what a query returned from a chosen location and device on the days they happened to run. There is no authoritative historical record of search rankings, and no tool can reconstruct a position nobody observed at the time.
Two habits fix most reporting. Segment branded and non-branded queries separately, because brand demand moves for reasons that have nothing to do with the work being done. And find a control — a section of the site that was not touched, or a comparable period — before attributing a change to anything. A decline that appears everywhere at once, including in areas the work never reached, is telling you something other than what the report says.
Where SEO programs fail
In my experience the failures are organizational far more often than they are technical.
- Tactics without a diagnosis. Someone buys content, or links, or a plugin, before establishing what is actually limiting the site. The most common finding in an audit is that the site's problem was not the one being funded.
- No owner with authority. SEO recommendations sit between marketing, engineering and whoever controls the platform. If nobody owns the queue, the audit becomes a document rather than a change.
- A content quota instead of a content strategy. Publishing to a monthly number produces pages that exist for the calendar rather than for a reader, and a large volume of those pages will drag on the whole site.
- Redesigns that ignore search entirely. A rebuild with no URL inventory and no redirect map is the fastest way to lose organic traffic that took a decade to build.
- Reporting theater. Rankings for terms nobody searches, presented monthly, in place of a plain statement of what changed and what it earned.
- Chasing the head term. One ambitious phrase absorbs a whole budget while the specific, lower-volume queries that convert go unwritten.
I will say plainly that some things are bad ideas: buying links, spinning content, hiding text, building a second site to rank twice, and paying for guaranteed rankings. They are bad ideas not because they are against the rules but because they stop working, and because unwinding them costs more than the results were ever worth.
How to work through this subject
If you are starting from nothing, start with access rather than tactics: a technical audit tells you what is currently possible, and it will usually change the plan. From there the two questions that most often decide competitive outcomes are the quality of the link profile and the clarity of the entity behind the site. The domain itself — the name, its extension, and its history — sits underneath all of it, which is why the domain material on this site is not a separate hobby but the same subject viewed from the infrastructure side.
This site is my professional record rather than a place to hire anyone. Consulting engagements are handled through Hartzer Consulting.