Service 03
A growing share of your category’s questions now get answered by an assistant that reads a handful of sources and names two or three brands. The work of being one of them is the same crawl-index-rank work you were already supposed to be doing, plus a few things most of the industry has wrong. We’ll build it into your site together, and into how your team publishes.
Google’s AI answers draw from the same index as the ten blue links, and a page has to be indexed and eligible for a snippet to be cited at all. There is no side door. No special markup, no AI text file, no submission form; Google says so in writing, and it says the “AEO hacks” being sold right now, chunking, FAQ schema, llms.txt, do nothing. What earns the citation is what always earned the ranking: a site the engine can read, and a page that says something the consensus didn’t already say.
That second part is where most content programs fail now. A clean, generic answer gets used by the assistant and never credited, because the model already had it. A page with a named point of view, a date, and something only you could have written gets named. So we build the technical foundation first, then make your site the source worth citing, and we measure it honestly, which means admitting what Google’s reporting can and can’t show.
The unglamorous half that decides whether anything else matters: what gets crawled, what gets indexed, what the rendered page actually contains, and how fast it loads. Most sites I audit are quietly hiding a third of their pages from the index and don’t know it.
Pages built the way the evidence says gets attributed: each section names its subject in the first sentence, answers directly, carries a visible date, and says something the consensus doesn’t. One thorough page serves many questions; a page per question is now a spam signal.
Which agents may read you, on what terms, and whether that trade is worth making, then enforced at the server rather than hoped for in a text file. Productized as a tool your team runs, built and documented here before the industry had a name for it.
Briefs built from the questions your buyers actually ask and the gaps in what currently ranks, written by people who know the subject, through a workflow with review gates. Google’s own guidance now draws the line: content a model could have produced is commodity, and the systems aren’t built to reward it.
Amazon’s search is its own engine with its own physics, and it now reads your off-Amazon marketing: branded search created elsewhere arrives at Amazon and pays out in rank. Site search and marketplace search run as one program. See Amazon & Marketplaces.
Multi-market search where the site, the catalog and the consent regime all differ by region, and hreflang is the single most commonly broken piece of the whole setup.
Crawl and index reality first, not a tool score. Which pages get fetched, which get dropped and why, what the rendered page contains, and which AI agents are already reading you. Search Console is the ground truth; everything else is an opinion.
Architecture, internal linking, speed, redirects, the schema that matters and none that doesn’t. This is where the compounding starts and the part most content-led programs skip. Your team learns the checks so they hold after I’m gone.
Pages shaped so a model can lift a clean, attributable claim. Then the truth about measurement: Search Console’s generative AI report shows impressions only, no queries, and nobody can see inside ChatGPT. We track what can be tracked, sample the rest on a schedule, and never present a survey as a census.
A crawl-and-answer audit tells you which pages are reachable, which agents are reading you, and where your brand currently shows up when an assistant answers for your category. Usually the first finding is a page Google dropped a year ago.
Everything you've been told about search is about four years out of date, and the people telling you are mostly selling something. Here is what SEO actually is in 2026, what keywords and meta tags and links actually do, how indexing and audits and tools actually work, and what it takes to be the source an AI assistant names instead of the one it quietly paraphrases. Written by someone who does this for consumer brands and has to live with the results.
Search engine optimization is the work of making a website easy for Google to find, understand and rank, so your pages show up in the unpaid results when someone searches for the thing you sell. That's the whole definition. SEO stands for search engine optimization; in your marketing org it's the organic search channel, the one you own instead of rent.
Ask your team how a search engine works and you'll get "the algorithm," said with a shrug. Here's the version you can actually run a business on. Google does three things, in order. It crawls: a bot follows links and fetches pages. It indexes: it decides what each page is about and files it. It ranks: for a given query, it picks which filed pages to show and in what order. Every SEO problem you will ever have lives in one of those three steps, and most of the expensive ones live in the first two, where nobody looks. A page Google never crawled cannot rank. A page Google crawled and judged thin, duplicate, or untrustworthy will not rank, and no amount of "content strategy" fixes a page Google has already decided not to keep.
Ranking is where the industry sells you mystery, so let me ruin the mystery. There are hundreds of signals and about five that move results: does the page actually satisfy what the searcher meant; do the words and the entities on it match that meaning; how authoritative is the site, which still mostly means who links to it and who mentions it; is the experience decent (fast, mobile, not buried under interstitials); and does the content show real experience and expertise. Google calls that last bar E-E-A-T, for experience, expertise, authoritativeness and trust. E-E-A-T is a rubric Google's human quality raters score pages against and a description of what the systems try to reward. It is not a ranking factor you can flip on, and anyone selling you "E-E-A-T optimization" as a line item is selling you the rubric back.
The page of results is called the SERP. In 2026 a SERP is not ten blue links. It's ads, shopping tiles, a map pack, videos, People Also Ask, and, on a growing share of queries, an AI Overview or an AI Mode answer written from the same index. "Search engine positioning" is the old name for where you sit on that page. The page got more crowded; the job didn't change.
What SEO is in your digital marketing mix is the channel that compounds. Paid search stops the second the budget does. A ranking you earned keeps producing after the invoice is paid, which is why the two run together: paid covers what organic can't reach yet, organic retires paid as it climbs. The cost of SEO is time. Nothing meaningful moves on a competitive query in less than months, and anyone promising weeks is describing paid search with the word "organic" stapled on.
SEO is crawl, index, rank. The money is almost always lost in the first two, where nobody's looking.
SEO is not dead. What died is the version where you ranked in ten blue links and collected the click. Since 2024 the same work has become the price of admission to AI answers, and the brands reading the obituaries are the ones about to be left out of them.
Here's what's actually happened. Google's AI Overviews and AI Mode build their answers from the same index that organic search uses. To be cited at all, a page has to be indexed and eligible for a snippet. There is no side door, no AI submission form, no special markup; Google says so in its own documentation. So "should we still do SEO" has become "do we want to exist in the set of pages the engine reads from," and the answer to that one writes itself.
The "SEO is dead" argument isn't stupid, it's just incomplete. AI answers do reduce clicks on some queries. An independent audit of 55,393 queries run over 40 days in March and April 2026 found an AI Overview on 13.7% of searches overall and on 64.7% of question-form searches. On those queries, fewer people click through, and that's real. But the same audit found that 29.8% of the domains the AI cited weren't in the top ten organic results at all. Read that again. The pool the engine draws from is wider than the old top ten. The game moved from "rank in the ten links" to "be in the set the engine reads," and the technical and content work that gets you into that set is the SEO work you were already supposed to be doing.
Whether it's worth it for your company turns on three questions, and I'd ask them in this order. Horizon: if you need results inside a quarter, SEO is the wrong tool and paid is the honest one; come back to this when you can fund something that pays out in year two. Demand: if nobody searches for what you sell, there is nothing to rank for, and your problem is demand creation, not search. Economics: organic traffic is not free; you pay for it in content, engineering and patience, so the question is whether a visitor from search is worth more than that over the life of the page. For a consumer brand with real search demand, the answer is yes, and the reason is the one that should decide every channel question: the asset keeps producing after you stop paying.
And yes, keywords still matter, but not the way the 2012 guide your agency is still quietly running from says they do. Engines don't match strings anymore; they match intent. Your page needs to be unmistakably about one thing a searcher wants, in the language searchers use, and better at it than the alternatives. Repeating a phrase does nothing. Being the clearest answer does everything.
SEO didn't die. It became the entry ticket to AI answers, and the ticket is the same crawl-index-rank work you've been underfunding.
Doing SEO on a website comes down to six things, and the order matters more than the effort: make sure every page that matters can be crawled and indexed; give each page one job; make the page do that job better than the competing pages; link your pages together on purpose; keep the site fast; earn references from other sites. Everything on every "47-point SEO checklist" is a sub-bullet of one of those six.
Here's the sequence, and I'd run it in this order because each step wastes money without the one before it.
robots.txt isn't blocking something important, that key pages aren't marked noindex, and that the Pages report says they're actually indexed. A shocking share of "why don't we rank" conversations end right here, with a developer's staging block that shipped to production two years ago.An SEO strategy is just the written version of those decisions: which queries, on which pages, in what order, and what number tells you it worked. "SEO research" is the first half of that, understanding what people search for and what currently ranks for it before you touch a page. When someone asks how to "increase SEO," the real answer is one of three levers and the Search Console data tells you which: get more pages properly indexed, make existing pages answer their query better, or earn more references. Pick the bottleneck. Don't do all three badly.
Six steps, in order. The order is the strategy; most teams skip the first two and spend the budget on the last one.
SEO gets split into four kinds of work, and the split is about where the work happens, not about different goals. Your agency's proposal will list them as four services. They're one program with four locations.
On-page is everything inside the page you control: title tag, headings, body copy, images and alt text, internal links, URL, structured data. This is what makes a page unmistakably about its subject. Off-page is everything that happens on other people's sites: links to you, mentions of your brand, reviews, citations in directories and the press. This is what makes your site trusted, and you can influence it but you cannot control it, which is why agencies love to sell it. Technical is the plumbing that lets the engine crawl and index you at all: architecture, crawlability, indexation controls, rendering (especially if your developers love JavaScript), speed, mobile, HTTPS, canonicals, hreflang, redirects, sitemaps. On a big site, technical problems silently erase the value of every dollar spent on the other three. Local is the version for a business that serves a geography: a complete Google Business Profile, the same name, address and phone number everywhere on the web, local reviews, pages that name the places you serve. The map pack runs on a partly separate system where distance to the searcher is a major factor, so a business can own its city and be invisible nationally, or the reverse.
The order a program should run them in: technical, then on-page, then off-page, with local layered in where the business is local. The most expensive mistake in this field is buying link building for a site that isn't properly indexed. It's like running ads for a store with the door locked.
Four locations, one program. Technical first, because a locked door makes every other dollar worthless.
A keyword is what somebody types into Google. That's it. The industry calls queries "keywords" for historical reasons, and a keyword can be one word, running shoes, or an entire question, what running shoes are best for flat feet. When your agency says "we're targeting 200 keywords," they mean "we've picked 200 searches we'd like your pages to show up for," which is a plan, not a result.
Here's how keywords actually work, because the folk version is still costing people money. When someone searches, the engine interprets what they meant, then pulls pages whose content, titles, links and context match that meaning. The keyword is the evidence the engine uses to connect your page to the searcher's intent. A page about trail running shoes that never says trail, running or shoes is hard to connect; a page that uses those words naturally in the title, the headings and the body is easy. That is the entire mechanism. There's no secret field.
What keywords do not do in 2026 is work by repetition or by exact order. Google has matched synonyms and variants for well over a decade and understands meaning well enough that running shoes flat feet and best shoes for running with flat feet return nearly the same page. Keywords do not have to be in a particular order. There is no density to hit. A page that says the phrase twenty times reads as spam to a human and to the machine, and the machine's opinion is the one that costs you. The discipline that works: pick one primary query per page, say it plainly in the title and the opening, then cover the subject well enough that the page also answers the fifty other ways people phrase the same question. "Keyword analysis," done properly, is figuring out what a pile of queries is really asking so one good page can answer all of them.
A keyword is a question. Your page either answers it better than the next page or it doesn't. Repeating the question is not an answer.
Keyword research is finding out what your customers actually type into Google, how often, what they mean by it, and which of those searches your site can realistically win. Done right it produces a short list of queries mapped to specific pages. Done the way it's usually done, it produces a spreadsheet of eleven hundred phrases that nobody opens again. I know because I've been handed that spreadsheet.
The method, for a website, a blog, or a whole content program:
The best keywords for your site are the ones where relevance, intent and a real shot at ranking line up, and that's almost never the head term. "Running shoes" at a million searches a month belongs to brands and retailers you will not displace this decade. "Running shoes for overpronation under $100" might be yours by spring. Volume is the number tools show you because it's the number that makes the tool look valuable. It is the least important column on the sheet.
Keyword cannibalization is what happens when step five is skipped and three pages on your site all target the same query. They compete with each other, Google can't tell which to rank, and usually none of them does well. The fix is to merge them into one page and redirect the rest, or to give each a genuinely different intent. Most sites older than three years have this problem and don't know it.
Research is collapsing a thousand phrasings into the hundred questions that matter and giving each one a page. The spreadsheet is not the deliverable.
Keywords go on a website in exactly the places a reader and a search engine look first: the title tag, the H1, the opening paragraph, the subheadings, the body where it reads naturally, the URL, the image alt text, and the anchor text of links pointing to the page. That's the complete list. There is no hidden field, no meta box, no plugin setting that makes a page rank, and anyone who tells you otherwise is about to sell you one.
For the one primary query a page is built for:
/running-shoes-flat-feet, not /p?id=8812.Keywords do not go in a meta keywords tag; Google has ignored it since 2009 and the field exists in your CMS the way a cigarette lighter exists in a new car. They do not go in hidden text, in text the same color as the background, or in a block of phrases at the foot of the page. Those are spam techniques, the penalties are real, and the people who sell them disappear before the penalty lands.
"Creating keywords for SEO" isn't a thing. You don't invent keywords; you find the ones people already use and decide which to go after. "Optimizing keywords" means the list above, then checking Search Console a few weeks later: if impressions for the query are rising but clicks aren't, the title or the opening needs work, because you're being shown and not chosen.
WordPress and Shopify put these fields in particular places; the platform specifics are on the web reference.
Eight places. If the keyword isn't in the title and the first sentence, nothing else on the page is going to save it.
A title tag is the HTML <title> element: the clickable headline of your search result and the text in the browser tab. It is the most influential on-page element for ranking and the first thing a searcher reads, so it should say what the page is about, lead with the main query, and stay under about 60 characters so it isn't cut off. Google rewrites titles it judges poor. A vague or stuffed title isn't just weak; it may be replaced by one Google wrote, and Google is not your copywriter.
The "SEO title" your WordPress plugin asks for and the "page title" Shopify asks for are the title tag. "Website title" usually means the same thing, or the site name that gets appended to it. The H1 is a different element, the visible headline on the page. They can match and usually should be close, but the title tag is what the SERP shows.
A meta description is the <meta name="description"> tag, a sentence or two that search engines may show under your title. It is not a ranking factor and hasn't been for years. Its only job is to earn the click, so write it as a specific promise of what the page delivers, around 150 to 160 characters, with the query in it naturally. Then accept that Google replaces meta descriptions with its own snippet from your page text on most queries, which means your opening lines matter at least as much as the tag you spent an hour on.
Meta keywords are dead, and have been dead for longer than most SEO agencies have existed. Google announced in September 2009 that it does not use the <meta name="keywords"> tag for ranking. Bing has said it treats the tag as, at most, a weak spam signal, meaning the only thing it can do is hurt you. How many keywords to put in a meta tag has a precise answer: none. Your CMS still shows the field because removing it would generate support tickets. Leave it empty.
So "using meta tags for SEO" in 2026 means three things that work: a title tag written for a human, a meta description written to be clicked, and a robots meta tag only when you deliberately want a page kept out of the index. Structured data gets lumped in here and isn't a meta tag; it's schema.org markup that can make a page eligible for rich results, and it does nothing special for AI answers beyond what Google already reads from the visible text, which Google has said in writing.
Title tag and opening sentence do the ranking. Meta description does the clicking. Meta keywords do nothing, and have since 2009.
A backlink is a link from someone else's website to yours. Search engines treat it as a reference: a site that many relevant, trusted sites point to is judged more authoritative than one nobody points to. Link building is the deliberate work of earning more of them, and it is the part of SEO with the highest ratio of money spent to results produced, because most of what's sold as link building is buying references from sites nobody trusts.
Backlinks still matter in 2026. Their value is just wildly uneven. One link from a genuinely authoritative, topically related site outweighs hundreds from directories, comment sections and the link farms your spam folder knows about, and the hundreds can get you penalized under Google's link spam policies. Paid links that pass ranking credit, link swaps at scale, automated schemes: all explicitly against policy, all still sold every day.
How you actually get backlinks is by giving other sites a reason to cite you, which is harder than buying them and is the only version that lasts:
Nofollow is a link attribute, rel="nofollow", that tells engines the linking site isn't vouching for the destination. Since September 2019 Google treats it as a hint rather than a command, alongside rel="sponsored" for paid links and rel="ugc" for user-generated content. Do nofollow links help SEO? Directly, assume not: a nofollow link typically passes little or no ranking credit, and that's the safe planning assumption. Indirectly, they can: a nofollow link from a major site still sends people, puts the brand in front of someone who may link later, and helps crawlers find pages. A natural link profile has both kinds. A profile with zero nofollow links looks exactly like what it is.
Links are references, and you earn references by being worth citing. Everything sold as "link building" that skips that step is buying a penalty on a payment plan.
To get a website indexed by Google, make sure the bot can reach it, tell Google it exists, and give it pages worth keeping. Most people do the second and skip the first and third, then wait.
The sequence:
robots.txt must not disallow what you want indexed. Pages must not carry a noindex tag or header. The site returns 200, isn't behind a login, and isn't a staging copy still wearing its old block. I'd check this before anything else because it's the cause more often than it has any right to be.For Bing, Bing Webmaster Tools does the same job and supports IndexNow, which lets your site ping participating engines the instant a URL changes. Google doesn't use IndexNow, so don't let anyone tell you it helps there.
How long it takes runs from hours to weeks. A new page on an established, frequently crawled site is often indexed within a day. A new domain with no links can wait a month. "Making Google crawl faster" is earning links, keeping the site fast and error-free, and publishing things that justify the crawl budget. It is not submitting the URL again.
Indexing is three gates: reachable, known, worth keeping. The third one is where the "we submitted it and nothing happened" pages are.
A canonical tag, <link rel="canonical" href="…">, tells search engines which URL is the official version of a page when several URLs show the same or nearly the same thing. Tracking parameters, HTTP versus HTTPS, trailing slashes, filtered category pages: every site generates duplicates, and the canonical consolidates their ranking signals onto one. "Canonical" means the standard version. Google treats the tag as a strong hint and usually honors it, and will overrule it when your own signals contradict it, for instance when every internal link points at the other URL. Fix the links, not the tag.
Hreflang tells engines which language and region each version of a page is for, so a searcher in France gets the French page and one in Canada gets Canadian English. It's set in link tags, HTTP headers or the sitemap; every version has to reference every other version including itself; and it is the single most commonly broken piece of international SEO, usually because someone did half of it.
Is PageRank still used? Yes. It's Google's original algorithm for scoring a page by the quantity and quality of links pointing at it, and Google engineers have confirmed repeatedly since 2017 that a form of it remains in the ranking systems. What's gone is Toolbar PageRank, the public zero-to-ten score, retired in 2016. You can't see the number. The idea underneath it, that links are votes weighted by the authority of the voter, is doing more work than ever.
Semantic SEO is optimizing for what a search means rather than the literal words: covering a topic's related concepts and entities, answering the questions around it, structuring content so a machine can tell what things are and how they relate. It's the obvious response to engines that read intent, and under a different name it's exactly what makes a page a good source for an AI answer.
Google Discover is the feed on the Google app and the Android home screen, chosen from a person's interests rather than a query. You can't opt in; indexed pages that meet the content policies are automatically eligible. What earns Discover traffic is timely, visually strong content from a site with demonstrated expertise, with large images (at least 1200 pixels wide, max-image-preview:large set) and a title that says what the page is rather than teasing it. Discover traffic arrives in floods and leaves without warning. Treat it as a bonus. Anyone building a plan on it is building on weather.
Canonicals fix duplicates, hreflang fixes languages, PageRank never left, and Discover is weather.
To check where you rank for a keyword, open Google Search Console. The Performance report shows every query your site appeared for, with impressions, clicks, click-through rate and average position, sliceable by page, country and device. It is the only source that reflects what Google actually showed people, and it answers "what keywords do I rank for" directly: filter to a page, read the list.
Googling yourself is a rough check at best. Results vary by location, device, history and the personalization Google applies even in an incognito window, so the position you see is not the position anyone else sees. For tracking a defined list of keywords over time, a rank tracker (Semrush, Ahrefs, SE Ranking, Moz, a dozen others) runs the searches from fixed locations on a schedule and charts the movement. Read the trend. A single day's number is noise, and the agency that reports a single day's number as a win knows that.
Checking another website's traffic is estimation, and I want to be blunt about this because a lot of strategy decks are built on it. No tool outside that site's own analytics can see its real visitor count. Similarweb, Semrush and Ahrefs produce estimates from panels, clickstream data and models built on rankings and search volumes. They are useful for comparing competitors against each other and for direction, and they are routinely wrong by a factor of two or more for any individual site. "How much traffic does this website get" has one honest answer unless the site is yours: nobody outside the building knows. For your own site, the number is in Google Analytics; see checking your own traffic.
The most visited websites in the world, by those same estimates in 2026, are Google, YouTube, Facebook, Instagram, WhatsApp, ChatGPT, Wikipedia, X and Reddit, with the order shifting by source. ChatGPT in that list is the clearest signal you'll get about where attention has moved.
Measuring SEO in Google Analytics means isolating the organic channel. In GA4, Traffic acquisition groups sessions by default channel; Organic Search is your line. Add landing page to see where organic visitors arrive, and key events to see what they did. Link Search Console to the GA4 property and the query and position data sits beside the behavior data, which is the view every honest SEO report is built from.
Search Console tells you the truth about your own rankings. Everything about someone else's traffic is a guess with a logo on it.
An SEO audit is a structured review of everything that affects a site's visibility: can it be crawled and indexed, what technical faults it carries, how well each page targets its query, the quality and structure of the content, the internal linking, the speed, and the links pointing in. The output is a ranked list of problems and fixes, ordered by what each one is worth. If the audit you were handed is a 90-page PDF with a score on the cover and no prioritization, you bought a tool export with a logo on it.
A real website audit needs three kinds of tool, and the first one's free:
The SEO software most teams actually use is a short list, and the list is shorter than the invoice: Search Console and Google Analytics (free), Keyword Planner (free with an Ads account), one of Semrush, Ahrefs or Moz for keyword, competitor and link data (from about $100 a month), Screaming Frog (free tier, paid for big sites), PageSpeed Insights (free). Everything beyond that is a specialist addition you add for a reason. A SERP API such as SerpApi pulls search results programmatically at scale; it's for people building rank trackers and research tools, not for site owners.
"SEO checkers" and "SEO scores," the sites that tell you your page scores 74 out of 100, grade a page against the tool's own checklist. Useful for catching a missing title or a four-megabyte hero image. Not a measure of how Google rates anything. A perfect score has no bearing on rankings, and Google has no SEO score to compare it to.
An SEO report should answer three questions with Search Console data: what changed (clicks, impressions, position, indexed pages, by month and against last year); why (which queries and pages drove it, what you shipped, what Google updated); and what's next. An SEO checklist for a new site is shorter than any you'll find published: verified in Search Console, sitemap submitted, no accidental noindex or robots block, one intent per page, title and H1 on every page, readable URLs, internal links to everything important, images compressed and described, mobile usable, HTTPS, Core Web Vitals passing, analytics installed. Twelve items. The 90-page version is the same twelve with padding.
Three tools, and the first is free. If the audit has a score on the cover and no priorities inside, it's an export, not an audit.
Content marketing is attracting and keeping customers by publishing things they find useful (articles, guides, tools, data, video) instead of interrupting them with ads. It works for a reason that hasn't changed: people search for answers before they search for products, and the brand that provided the answer gets the trust and the visit. SEO and content marketing overlap almost completely. Content is what ranks; search is how most content gets found; a content program without search is a newsletter, and a search program without content is a technical audit with nowhere to go.
A content strategy is the plan behind the publishing: who it's for, what questions and topics it covers, in what format, how it gets found, which commercial goal it serves, and how you'll know. Without one, content marketing becomes a calendar of posts nobody asked for, which describes most corporate blogs. With one, every piece has a query, an audience and a job.
A content audit is an inventory of everything already published: every URL with its traffic, rankings, links, conversions and age, and a decision for each: keep, update, merge, or remove. Run one on a site older than three years and the same thing shows up every time: a handful of pages produce nearly all the value and a long tail of thin, overlapping posts is dragging the whole site down. Pruning and merging that tail is regularly the single most effective SEO action available, and it's the one nobody wants to do because it means admitting the calendar was the problem.
Writing SEO content means writing for a reader who arrived from a specific search. Answer the query in the opening, then go as deep as the question deserves, use the searcher's own words in the title and headings, name things precisely, and have it written by someone who actually knows the subject. Google's guidance updated in July 2026 draws the line in plain language: content that recycles what the internet already says, or that a generative model could have produced just as easily, is commodity content, and the systems are not built to reward it. Their own example is instructive. Commodity: "7 Tips for First-Time Homebuyers." Not commodity: "Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line." First-hand experience, your own data, a real position. That's what distinguishes a page now, and it's what most content programs are structurally unable to produce because they were built to hit a cadence.
B2B content marketing follows the same rules with a slower buyer: fewer, deeper pieces aimed at the questions a committee asks over months, distributed where those buyers actually read, which is almost never your blog alone. Keyword research for content marketing is ordinary keyword research with the weight on informational queries, the "what is," "how to" and "why" searches, grouped so one piece answers the whole cluster instead of one phrasing.
Content that could have been written by anyone will be ignored like it was written by no one. Google now says this in its own documentation.
Can you track your brand's mentions in AI answers? Partly, and anyone selling it as complete is overstating what can be seen. Here is exactly what exists as of October 2026, because this is the question I get asked most and the one with the most confident wrong answers in circulation.
Google AI Overviews and AI Mode. Search Console has a Generative AI performance report, rolled out to every site on 31 August 2026. It shows impressions, meaning how many times links to your site appeared inside a generative AI feature, by page, country, date and device. It does not show clicks, click-through rate, position, or, the part that matters, the queries. AI Overviews and AI Mode are reported together with no way to separate them. The report doesn't appear at all until you have enough impressions. So you can learn that your pages were shown in AI answers and nothing about what was asked or what came of it. That is the entire first-party measurement Google offers, and it's more than anyone else offers.
ChatGPT. OpenAI gives brands no reporting on how often or in what context they're mentioned. None. What you can see is referral traffic: visits from ChatGPT arrive with chatgpt.com as the referrer, and links ChatGPT includes in answers carry utm_source=chatgpt.com, so they show up as a source in GA4. That measures clicks out, not mentions.
Gemini. Same. Referrals show as gemini.google.com; there is no mention reporting.
Beyond that, the only method is sampling: run a fixed panel of prompts your buyers plausibly ask, on a schedule, across engines, and log whether and how the brand appears. Tools like Profound, Peec AI, Semrush's AI toolkit and Ahrefs' Brand Radar automate this at scale. What they produce is a trend line and a competitive comparison, which is useful. What they do not produce is a census, because every answer is generated fresh and varies by user, session and phrasing. Treat the output as a survey. Any vendor presenting it as a report is grading their own homework.
Being found in AI search in the first place rests on ordinary SEO plus a few things worth knowing. Google's own guidance, updated July 2026, says there are no additional requirements to appear in AI Overviews or AI Mode: a page must be indexed and eligible to be shown with a snippet, and that's it. No special markup. No AI text file. No submission. Google Search ignores llms.txt; maintaining one, in Google's words, "will neither harm nor help." FAQ rich results were deprecated from Google Search on 7 May 2026 and the documentation removed in June, so FAQ schema isn't a route in either, whatever the plugin says. What Google does recommend is unglamorous: content in text, findable through internal links, with a point of view that isn't a recycled one.
Two mechanics are worth understanding because they change how you build pages. Both features use query fan-out: the system fires several related sub-queries across sub-topics and assembles the answer from the best passage for each. That's why one thorough page can get drawn on for many questions, and it's why Google explicitly says not to build a page per query variation, which it now treats as scaled content abuse. And sources are chosen at the passage level, not the page level, which rewards pages where each section names its subject in the first sentence, answers directly, and carries a visible date. The independent evidence on what gets cited versus what gets silently absorbed points the same way: passages with a named entity, a current date, and something the consensus didn't already say are the ones that get attributed. The clean consensus paragraph gets used and not credited. That's the trap most "AEO" content walks straight into.
You can see that you were shown in AI answers. You cannot see what was asked. Anyone selling you more than that is selling a survey as a census.
The best way to learn SEO is to read Google's own documentation and then apply it to a site you control. Google Search Central publishes the SEO Starter Guide, the full technical documentation, the spam policies and, since 2025, explicit guidance on generative AI features. It's free, current and primary, and most paid courses are a paraphrase of it with a certificate attached.
For structure, the free academies from Semrush, Ahrefs and HubSpot cover the fundamentals properly, and Moz's Beginner's Guide to SEO is still the right first read. Google offers no official SEO certification. Anything marketed as a "Google SEO certification" is somebody else's. The Google Analytics and Google Ads certifications are real and adjacent; they certify those products, not SEO.
Past the basics, SEO is learned by doing, and there's no shortcut I've found. Put Search Console on a site, publish pages against a keyword plan, watch what happens for six months, read the data until it makes sense. That experience is what clients pay for, and it cannot be acquired from a course because the course doesn't have your data in it.
For SEO news, the sources practitioners actually read: Google's Search Central Blog and its ranking updates page for anything official; Search Engine Land, Search Engine Journal and Search Engine Roundtable for daily coverage; the research blogs at Semrush, Ahrefs and Moz. Barry Schwartz's Search Engine Roundtable is where algorithm-update chatter surfaces first. Treat any "leaked ranking factor" as entertainment until Google's documentation or a reproducible test confirms it, which it almost never does.
Google's docs plus a site you can break. The certificate industry exists because that's less fun to sell.
SEO services are the work in this reference, done by someone outside your building: a consultant (one person), an agency (a team), or a freelancer. The services themselves are the audit, the technical fixes, the keyword research, the content plan and production, the link earning, and the reporting. A "package" is some bundle of those on a monthly retainer.
What a good provider does first is tell you what's actually wrong, with evidence from a crawl of your site and from your Search Console, before proposing anything. What a bad one does first is promise a position. Google's own guidance on hiring an SEO is as blunt as I'd be: nobody can guarantee a number-one ranking, and a provider who does is a warning sign, as are unsolicited emails, "secret" techniques, and anything involving link schemes. I would add: an agency whose monthly report could be forwarded unedited as your SEO strategy doesn't have a strategy. It has an invoice with charts.
What it costs in the US in 2026: consultants typically bill $100 to $300 an hour or a fixed project fee; a one-time audit of a mid-size site runs a few thousand dollars; agency retainers commonly run $2,000 to $10,000 a month for small and mid-market businesses and well past that for enterprise. The $299-a-month packages buy automated reports and low-quality links, which is worse than spending nothing, because the links can cost you a penalty on top of the fee.
Local SEO services are the subset for a business serving a geography: Google Business Profile management, citation consistency, review strategy, location pages. Searching "SEO services near me" is reasonable for local SEO, where knowing the market helps. For everything else, the provider's address is irrelevant; the work happens in Search Console and a crawler, and the right consultant for a consumer brand is the one who has done it for consumer brands, wherever they sit.
Three things to ask for before signing anything: findings from a crawl of your own site, before the contract; a client they've worked with for more than a year; and a description of how they report, which should center on Search Console clicks and impressions and on your revenue, not on a tool's visibility score. If any of the three produces a pause, you have your answer.
Ask for the crawl before the contract. A provider who can't tell you what's wrong before you pay won't tell you after, either.