What is SEO — and how do you actually improve it?

How search engines actually work

Every ranking you've ever seen is the output of three machine steps, and every SEO tactic that works maps to one of them:

1 · Crawling. Software agents ("crawlers" — Googlebot, Bingbot) follow links from page to page, downloading what they find. If a page can't be reached by links, is blocked by your robots.txt file, or only exists after heavy scripts run, it may simply never be collected. Crawling is a budgeted activity — engines spend more of it on sites that are fast, well-linked, and frequently updated.

2 · Indexing. The engine parses each collected page and files it in an enormous library keyed by meaning: what the page is about, what questions it answers, how it relates to your other pages, whether it duplicates something else. This is where your titles, headings, structured data, and internal links do their quiet work — they're the labels the librarian reads. A page can be crawled and still not indexed if the engine judges it thin, duplicated, or unimportant.

3 · Ranking. When someone searches, the engine pulls candidates from the index and orders them using hundreds of signals — relevance to the words and the intent behind them, the page's demonstrated authority, its speed and usability on the searcher's device, its freshness, and the engine's confidence that the source is who it claims to be. Ranking happens per-query, per-location, often per-person. There is no single "your ranking."

Hold onto that model. Every improvement in the rest of this guide is just making one of those three steps easier for the machine — and, not coincidentally, better for the human behind the search.

What search engines reward in 2026

Ranking systems change constantly in the details and slowly in the fundamentals. The durable rewards, across twenty years of updates, are these:

• Satisfying the search, quickly. The page that answers the question — visibly, near the top, without making the visitor dig — beats the page that ranks for the phrase but hides the answer. Engines measure satisfaction indirectly but relentlessly.
• Demonstrated experience and expertise. Google's own quality framework (E-E-A-T: experience, expertise, authoritativeness, trustworthiness) favors content that shows first-hand knowledge — real specifics, real process, real authorship — over generic filler. This matters more, not less, now that generic filler is free to generate.
• Technical legibility. Real text in real HTML, clean heading structure, working links, one canonical version of each page, structured data that matches the visible content. Machines can't reward what they can't parse.
• Honest speed on phones. Core Web Vitals — how fast the main content appears (LCP), how quickly the page responds to taps (INP), how much the layout jumps around (CLS) — are measured on real users' devices and folded into ranking. Most sites' worst scores are on the phones most of their customers use.
• Corroborated identity. A business whose name, offering, location, and claims are consistent across its site, its business profiles, its reviews, and the wider web is safer to rank than one that contradicts itself. Trust is triangulated.

Improving technical SEO — the machine-legibility layer

Technical SEO is everything that determines whether engines can crawl, render, and correctly file your site. It's the least glamorous pillar and the one most often silently broken.

Make every page reachable and unambiguous

• Internal links are the roads. Every important page should be reachable within about three clicks of your homepage, through real HTML links (not script-only buttons). Orphan pages — reachable only by knowing the URL — rank poorly because they're barely crawled and receive no internal authority.
• One address per page. If the same content answers at multiple URLs (with and without www, with tracking parameters, in near-duplicate variants), engines split its credit. A canonical tag on each page declares the one true address; redirects retire the rest.
• Sitemap and robots.txt, current and honest. The XML sitemap is your table of contents for crawlers — it should list exactly your real, indexable pages and nothing else. robots.txt should block only what you genuinely want unread.

Make pages render as text, not as a puzzle

• Content in HTML first. If your text only appears after JavaScript executes, some crawlers see it late and some AI crawlers never see it at all. Server-rendered or static HTML with scripts as enhancement is the robust pattern — it's how this site is built.
• Headings as an outline. One h1 stating what the page is; h2/h3 sections a machine could read as a table of contents. If your headings, read alone, don't summarize the page, restructure.
• Structured data (schema). JSON-LD markup — Organization, Service, FAQPage, Article, BreadcrumbList — tells engines in their own vocabulary what your page claims. It must mirror visible content exactly; markup that says what the page doesn't is a trust liability, not an asset.

Make it fast where it counts

• Images are the usual culprit. Serve them sized to their containers, compressed in modern formats, lazy-loaded below the fold, with dimensions declared so the layout doesn't jump.
• Fonts and scripts second. Preconnect to font hosts, load non-critical scripts deferred, and let the first paint be the page's actual content — not a spinner.
• Measure on real phones. Lab tests on a fast laptop flatter you. Field data — what Google actually uses — comes from your visitors' devices. A mid-tier phone on mediocre signal is the honest test bench.
• HTTPS, everywhere, without warnings. A security warning ends the visit and the ranking conversation simultaneously. (We wrote up what the "not secure" warning means separately.)

Improving content & on-page SEO — the answer layer

Content is where rankings are actually won, because it's where searches are actually satisfied.

Start from questions, not keywords

Modern engines match intent, not strings. So the working unit of SEO content is the question a real customer asks. List the twenty questions you get asked in sales conversations, in your inbox, on the phone. Each is a page or a section. The phrasing your customers use — "why is my website so slow," "how much does a custom site cost" — is the phrasing to put in titles and headings, because it's the phrasing engines are matching against.

Answer first, then earn the depth

Structure every substantive page the way this one is structured: the direct answer in the first two sentences, visibly boxed or bolded, then the detail below for readers who want it. This "answer-first" pattern wins three audiences at once: skimming humans, search engines choosing featured snippets, and AI assistants choosing what to quote. Burying the answer under scene-setting loses all three.

On-page mechanics that still matter

• Title tag: the page's question or promise, in the customer's words, unique across your site, roughly 50–60 characters so it doesn't truncate. It's both a ranking input and your ad in the results list.
• Meta description: not a direct ranking factor, but it's your pitch for the click — a sentence of genuine promise beats a keyword pile.
• Internal links with meaningful anchors. "Read our guide to why websites are slow" passes topical meaning; "click here" passes nothing. Every new page should link to, and be linked from, its topical neighbors — that's how engines learn your site's structure of expertise.
• Freshness with receipts. Visible updated-dates on maintained pages, and real revisions when facts change. Engines increasingly prefer demonstrably current sources; so do AI answers.
• Depth over duplication. One thorough page per topic beats five thin overlapping ones — thin variants compete with each other and dilute authority. Consolidate; redirect the losers.

Show experience — the part AI content farms can't fake

Generic explainer text is now free, so it's worth approximately nothing. What ranks — and converts — is evidence of first-hand work: your actual process, your measured results, your named tools, your real photographs, your opinionated recommendations with reasoning. Write what only you can write; that's the moat.

Improving authority & off-page SEO — the trust layer

Engines corroborate. A site that exists in a vacuum — no links pointing to it, no mentions, no reviews, no profile presence — is hard to trust no matter how good its pages are. Building authority honestly:

• Earn links by being citable. The durable link strategy is publishing things worth pointing at: original data, genuinely useful tools, definitive answers, strong opinions well argued. One relevant link from a real industry site outweighs dozens of directory-spam links — and purchased link schemes are a documented way to lose rankings, not gain them.
• Be present where your industry talks. Trade associations, local business organizations, supplier and partner pages, event listings — unglamorous, legitimate, and cumulative.
• Reviews are authority. A steady stream of genuine reviews on Google and industry platforms — asked for honestly, responded to professionally — feeds both local rankings and the AI answers that summarize "what people say about" businesses like yours.
• Consistency everywhere. Identical business name, description, address, and phone across your site footer, Google Business Profile, directories, and socials. Every stale variant you fix is a contradiction removed from the record.

Local SEO, if you serve an area

If customers choose you partly by geography, a second ranking system applies — the map results — and it runs on: a complete, actively maintained Google Business Profile (correct categories, services, hours, photos, posts); review velocity and responses; consistent citations of your name/address/phone across the local web; and location-relevant pages on your site that say, in real text, where you work and what you do there. Local intent queries ("web design studio Oregon") blend both systems, so everything in this guide feeds the map too.

Measuring whether it's working

SEO without measurement is astrology. The honest instrument panel:

• Google Search Console (free) — the ground truth for which queries show your pages, your average position, what gets clicked, and what's failing to index. Check the Pages and Performance reports monthly at minimum.
• Core Web Vitals field data — in Search Console and PageSpeed Insights — for the speed your real visitors experience.
• Rankings as trend, not scoreboard. Positions vary by device, location, and day. Watch the 3-month trend for your priority queries, not the daily wobble.
• Business outcomes. Calls, form fills, sales from organic visits. Rankings that don't move revenue are a hobby.

The honest timeline: technical fixes register in days-to-weeks after recrawl; content improvements in one to three months; authority over quarters. It compounds — the curve is slow, then suddenly not. We wrote a fuller version of this in how to rank higher on Google.

The first 30 days, step by step

1. Get a baseline reading. Measure before you touch anything — speed, indexing status, current queries in Search Console, and a graded pass over structure, content, and trust signals. (This is literally what The Grader does; use it or assemble the same picture by hand.)
2. Fix the blockers. Anything preventing crawling or rendering: broken pages, redirect chains, script-dependent content, missing canonicals, security warnings. Machine-legibility first.
3. Rewrite your five most important titles and openings. Your homepage plus your four highest-value pages: customer-language titles, answer-first openings, honest meta descriptions.
4. Publish two real answers. Two pages, each thoroughly answering one question customers actually ask, structured answer-first with proper headings and FAQ markup.
5. Lock identity down. Google Business Profile complete; name, description, and contact identical everywhere you appear; obvious stale listings fixed.
6. Compress the images. Usually the single largest speed win available in an afternoon.
7. Set the measurement habit. Search Console verified, a monthly 30-minute review on the calendar, and your priority queries listed so next month's comparison is honest.

What not to do

• Don't buy links or "guaranteed page one." Link schemes are explicitly penalized, and nobody outside a search engine can guarantee its results. Treat the promise as a vendor-quality signal — a bad one.
• Don't stuff keywords. Density is a 2009 metric. Repetitive, unnatural phrasing reads as spam to engines and humans alike.
• Don't mass-generate thin AI pages. Engines have been demoting scaled low-value content aggressively. Volume without experience is a liability.
• Don't chase algorithms. Sites built around update-chasing tricks get whiplash at every update. Sites built around answering customers well tend to benefit from updates.
• Don't set and forget. SEO is maintenance, like the books or the storefront. Quarterly attention beats an annual panic.

Where AEO and GEO fit in

An increasing share of searches now end in an AI-written answer rather than a clicked link — in ChatGPT, Perplexity, Copilot, or Google's own AI Overviews. Optimizing for those systems is called AEO (answer engine optimization) and, near-synonymously, GEO (generative engine optimization). The good news: they reward the same fundamentals this guide teaches — answer-first structure, real text, consistent identity, corroborated authority — plus a few specifics of their own. We've written both up at the same depth as this page — What is AEO?, What is GEO?, and the side-by-side SEO vs. AEO vs. GEO comparison — plus surface-specific playbooks for ChatGPT and Google AI Overviews. We grade all of it as one Findability lens — the commercial version of that reading is our SEO & AEO audit.

Common questions