Search everywhere optimization

Search everywhere optimization is the practice of improving your visibility on the search and discovery surfaces your audience actually uses, rather than on traditional search engines alone. That is the definition, and it is close to the wording Michigan Tech, Semrush and Ahrefs all arrive at independently.

Now the part almost everybody gets wrong. The important word in that sentence is not "everywhere". It is your audience. Read it with the emphasis on "everywhere" and you get a plan to be on eight platforms. Read it with the emphasis on "your audience" and you get a plan to be excellent on three and absent from the rest, which is the only version that works with a real budget.

The opponent in this chapter is the version of this idea that arrives as a trends article: search has fragmented, so you must now be present on Google, YouTube, TikTok, Reddit, LinkedIn, Pinterest, ChatGPT, Instagram and the app stores. That is not a strategy. It is a shopping list, and it is how a small team spends a year producing nothing on six surfaces.

The term has an author, which is worth knowing because an unattributed framework is a worse framework. Ashley Liddell at Deviation coined it a few years ago, spelled the British way. Rand Fishkin popularized it at SparkToro and says so explicitly, crediting Liddell for the original use and Nikki Lam, then head of SEO at NP Digital, for bringing it to his attention. Fishkin's post arguing that the industry should stop inventing acronyms and just use this one is dated 30 May 2025.

Fishkin also has the best one-line version of the underlying claim, and I am going to lean on it for the rest of this page: search is a behavior, not a channel. Once you hold that, the question stops being "which platforms are search engines now" and becomes "where do my buyers go when they have this problem". Those produce completely different plans.

Six surfaces will answer the same buying question six different ways, because each one is doing a different job for the person asking. That is the finding, and it takes about eight minutes to reproduce on your own screen.

The opponent here is the habit rather than a belief. Nobody argues that Google is the only search engine. But almost everybody still plans as though it is, because that is where the reporting is, and a channel with a report beats a channel without one in every meeting.

So run the experiment before you read my conclusions about it. Take one real buying question, ideally in a category you know nothing about, and put it through all six.

The useful question afterwards is not "how do I optimize for all of these". It is why did I use each one. Google, to get options and sources. The assistant, to have the reading done and the list shortened. YouTube, to see the thing. Reddit, because I did not believe the polished version. The review site, to check the shortlist was not a mistake. The website last, for the official facts and the checkout.

Every one of those is a different job. Which means a platform is not a strategy, and "we should be on TikTok" is not a sentence with a mechanism in it.

One precision point, because the popular version of this argument is sloppy and I would rather you had the defensible one. I am not saying YouTube, Reddit, ChatGPT and Google work the same way. They absolutely do not. YouTube says outright that its results are not a list of most-viewed videos. Google names relevance, distance and prominence for local results. Apple names text relevance across an app's name, subtitle, keywords and category, plus download and rating behavior. Four different retrieval systems with four different published rules. The claim that survives contact with any of that is narrower: your customer can use all of them to satisfy a search need.

People search for five reasons: to discover what exists, to understand how something works, to compare their options, to validate a choice, and to act on it. Get those five and you can place any query, any platform and any content idea without a list.

The opponent is the platform taxonomy, and I want to be fair to it because I used every one of them as research. Michigan Tech runs eight categories, including traditional search, app stores, ecommerce, social and video, AI, voice, local and niche platforms. Search Engine Land names seven platforms. Ahrefs concentrates on AI assistants, YouTube, Reddit, TikTok and Pinterest. All three are good reference material and none of them is teachable, because a beginner handed a list of platforms memorizes the list and learns nothing transferable. Then a new app ships and the list is wrong.

The fourth job is the reason this framework has five boxes rather than the three a 2019 funnel diagram would have given you. Validate used to be a step you could mostly satisfy with testimonials on your own site. It has become much harder to satisfy that way, because a buyer can now check a brand across reviews, communities, search results and AI answers before acting. Somebody hears your name in an AI answer, and their next three searches are your brand plus reviews, your brand plus reddit, your brand plus alternatives. I will come back to why that matters in chapter ten.

Those five jobs and the six surfaces in chapter five are the two halves of what I call 5 Jobs, 6 Surfaces, and the grid that pairs them is the decision instrument the rest of this page runs on. Being precise about what is mine in that: the pairing. Treating jobs and surfaces as two axes of one grid, and using that grid as the decision instrument rather than as a description, is my synthesis and the part I will defend. Everything else in it is borrowed and credited. Thinking in jobs rather than demographics is jobs-to-be-done, which is decades old and not an SEO idea. "Search is a behavior, not a channel" is Rand Fishkin's line. Compressing a long platform list into a handful of categories is what Michigan Tech, Search Engine Land and Ahrefs each did before me, at eight, seven and five. Use the framework however you like. There is no trademark on it and no gate in front of it.

Journeys are not linear, and any diagram that implies they are is lying to make itself tidy. People bounce. They validate before they compare, discover something new during validation, and abandon the whole thing for three weeks. The sequence in the diagram above is the common shape, not a funnel.

The fifth job is also the one moving fastest, and not in the direction the surface table further down implies. Watch what happens when the act job arrives on a surface that used to only do discovery.

Search really does happen in about twenty-three places, and Google is still nearly three quarters of it. Both halves of that sentence come from the same study, and holding both is the difference between a strategy and a trend take.

This chapter has two opponents and I want to knock them both down with one data set. The first is "Google is dead, stop doing SEO". The second is "AI search is hype, nothing has changed". They are both wrong, and they are wrong by different amounts.

Here is the only actual measurement I found while researching this page. Everything else I read about search everywhere optimization cited another article that cited another article. Rand Fishkin and Datos, which is a Semrush company, measured search activity across 41 domains on a desktop clickstream panel for all of 2025.

Read the AI row carefully, because it is where most 2026 strategy decks go wrong. Every AI tool combined came to 3.2% of searches in that panel. Commerce sites came to about 10%. Amazon, Bing and YouTube each recorded more desktop search activity than ChatGPT. If you have spent this year worrying about AI visibility and have never once looked at your presence in Amazon search or YouTube search, you have prioritized by news coverage rather than by where your buyers are.

And read the Google row carefully too, because the same study says its share fell 3.5 points across the year while the 34 sites outside the top seven grew theirs. That is the actual shape of the change: not a collapse, a slow leak into a long tail. The leak is the opportunity. The 73.7% is the reason you do not abandon the surface it is leaking from.

One more number, from the same author and a different study, because it is the one that reframes what a search is worth. In the first four months of 2026, 68.01% of Google searches ended without a click. Fewer than one in three sends traffic anywhere. So "impressions went up and clicks did not" is not a failure of your page. It is the current shape of the surface, and any plan that only counts clicks is counting a third of the thing.

I do not want to oversell one panel. Desktop only, US, 41 editorially chosen domains, and every search weighted equally whether it was a word in a box or a forty-turn conversation. Fishkin says as much himself and estimates Google could be as low as 65% with mobile included. It is the best public measurement of this question that exists. It is not the last word, and the caveats are inside the figure above rather than down here for a reason.

Six categories cover every platform anybody will name at you. Web search, AI search, video search, community search, commerce and review search, and local and app search. Learn six and you never need the list again.

Each row below carries what the platform itself says about how its search works, quoted and sourced. That is deliberate, and it is the part worth trusting in a year. What an SEO blog says about the TikTok algorithm is folklore with a publication date. What YouTube says about YouTube ranking is a primary source, and the useful thing about primary sources on this subject is how often they contradict the folklore.

Now the same six against the five jobs, because the pairing is where the decisions come from. A surface is not good or bad. It is good at some jobs and useless at others, and a business whose buyers do a lot of validating needs a completely different mix from one whose buyers mostly discover and act.

That grid is my judgement, not a measurement, and it says so inside the figure. If you disagree with a cell for your own market, you are probably right about your own market. The value is in noticing the shape: your own website is weak at four of the five jobs and indispensable at the fifth, and no amount of homepage work changes that.

No, you do not need to optimize for all six surfaces. You need two or three, chosen because your buyers use them, and you need to be genuinely good on those rather than present on all of them.

This is the practical rule I would put on the wall, and it is the one the sources agree on most strongly. Michigan Tech states it directly: "everywhere" does not mean literally everywhere, and you do not need to chase every new platform. Semrush flips the sequence entirely, telling you to interview recent customers about where they actually directed their curiosity and build presence there, rather than picking platforms from a trends list. Ahrefs gives the conditional version: complex, high-ticket products lean toward Google, YouTube and LinkedIn, while visual products skew toward Instagram, Pinterest and TikTok.

The mechanism is boring and it is the reason this rule holds. Every surface has a minimum viable investment before it returns anything, and that minimum is set by the competition on that surface rather than by your ambition. Half a presence on six surfaces clears the bar on none of them. There is no partial credit for a YouTube channel with four videos on it.

I have watched this go wrong in the same way at least a dozen times. A team reads a search everywhere article, adds five channels to the plan, and eleven months later has a dead TikTok account, a Pinterest board nobody has opened, an abandoned subreddit presence, and the same Google traffic they started with. The channels were not the mistake. Adding five at once was.

So the honest version of the advice is uncomfortable and short. Pick the surfaces your buyers use, name the ones you are deliberately skipping, and write the skip list down where your team can see it. There is a tool for exactly that in chapter twelve.

The same lesson produces four different answers for four different businesses, which is why a universal search everywhere strategy is bad advice to at least three of them.

The opponent is the recipe, and I include myself in the accusation. Anyone writing about this subject writes from their own market, and most of us are in B2B software, so most of the advice assumes a long research cycle with a validate step in the middle. Apply that shape to an impulse purchase or an emergency plumber and it is not slightly off. It is backwards.

Look at the first surface in each one. An assistant, then Maps, then TikTok, then a general search. Four different starting surfaces, which means four different discovery investments, which means the phrase "start with SEO" is only correct for one of them and only if you read SEO broadly.

The one thing all four share is the last two steps. Every journey ends on something you own or something you control the content of, and every journey has a brand search somewhere near the end. That convergence is the subject of chapter ten, and it is the reason the website did not stop mattering when discovery moved.

The first of those four is the shape I work in most, and I have written the channel version of it separately in SaaS SEO. Read this chapter for the journey and that page for what you actually build against it.

Four archetypes are useful and none of them is anybody's real data, so here is one that is. This is a property I own, over sixteen months, with the exports published.

Search everywhere optimization is a framework, and the thing it organizes is two lists that keep getting mixed together. Search surfaces are places people go. Optimization disciplines are work you do. SEO is not one of the six surfaces. Web search is the surface, and SEO is the work that wins it.

The opponent is the idea that there are now several competing disciplines and you are behind on the new one. There are six surfaces. The disciplines aimed at them overlap heavily, and most of what is sold as new is one of them with a different label on it. I say that as somebody who spends most of his working week on the AI surface specifically.

Google's guidance addresses the acronyms in one paragraph and then declines to adopt any of them. It defines AEO as answer engine optimization and GEO as generative engine optimization, notes they describe work focused on visibility in AI search experiences, and concludes that from Google Search's perspective optimizing for generative AI search is optimizing for the search experience, and thus still SEO. It then points you at its guidance on evaluating third-party SEO advice, which is not subtle.

I would still use the words, and here is the distinction I hold. SEO is the work of being discoverable, understandable and worth selecting, and web search is where most of it lands. AI SEO is the part of that work aimed at a surface where the output is a synthesis rather than a list, which changes what you write and completely changes what you measure. GEO and AEO are those same activities with a different label, usually attached to a product.

One qualifier, because this is where the popular version overreaches and I do not want to hand you a claim that falls over. AI search systems do not share one index. Google says its core Search ranking systems and its Search index are foundational to AI Overviews and AI Mode, and that is a statement about Google. OpenAI runs a separate crawler, OAI-SearchBot, specifically for whether you appear in ChatGPT search answers, and Perplexity runs PerplexityBot for its own. Different systems, different retrieval, different ranking, different citation behavior. What they share is the foundations: crawlable pages, useful content, verifiable authority and information a machine can actually read.

So the practical test for anyone selling you GEO as a separate service with a separate retainer is unchanged. Ask which of their deliverables would not also improve your web search visibility. If the answer is "none", you are paying twice for one thing.

Do not memorize any of this yet. I care much more that you understand the behavior, and the acronym you use for it changes nothing about what a buyer does with six tabs open.

If you want the deeper version of the mechanism underneath the AI surface, I wrote up why query fan-out is a passage problem rather than a page-length problem in query fan-out in SEO.

A grounded AI answer is assembled from pages a system retrieved, which is why the AI surface and the web surface cannot be worked separately. That is not a defense of SEO. It is the mechanism, and for its own generative features Google publishes it.

Google describes two techniques behind its generative features. The first is retrieval-augmented generation, which it also calls grounding: the system relies on the core Search ranking systems to retrieve relevant, current pages from the index, reviews the specific information in those pages, and then shows clickable links to the pages that support the response. The second is query fan-out, a set of concurrent related queries the model generates to fetch more than the original question would return. Google's own worked example is a lawn full of weeds fanning out to herbicides, chemical-free removal and prevention.

Read the first one again slowly. The generative layer is reading an index. If your page is not in that index, or is in it and not competitive for the sub-questions, there is nothing for the model to ground on. Which is why Google can say, and does, that there are no additional requirements to appear in AI Overviews or AI Mode, and no additional technical requirements beyond being indexed and eligible to show with a snippet.

Other assistants are not running on Google's index, so do not carry that sentence across unchanged. What does carry across is the shape: every one of them retrieves from something, every one of them publishes a crawler you can allow or block, and none of them can cite a page it could not fetch. OpenAI is explicit that OAI-SearchBot governs ChatGPT search visibility and that GPTBot, which covers training, is a separate decision.

There is one warning attached to that insight, and it is the mistake the insight causes. The answer is not a page per sub-question. Google names creating separate content for every possible variation of how people might search, including fan-out queries, as a scaled content abuse risk, and says plainly that it is an ineffective long-term strategy because a high quantity of pages does not make a site higher quality. Cover the question properly, in as few pages as it honestly takes.

Your website is the only surface where you set the facts. Everywhere else, you are being described, and in 2026 an AI answer about you is a synthesis of those descriptions.

The opponent is a real and reasonable conclusion that people draw from everything above: if discovery has moved off my site, why am I still maintaining a site. It is the right question with the wrong answer, and the mechanism in the previous chapter is why. Grounding means the model retrieves pages and reviews their specific information. Your pages are the specific information.

Which leads to the distinction that makes the validate job make sense without teaching a single new acronym. Everything you publish sits in one of three tiers of control, and how much of it is yours changes completely between them.

Now the sentence this chapter exists for. "We said we are the best on our homepage" and "independent sources keep describing us that way" are not the same claim, and only one of them survives a buyer doing the validate job. It is also the difference between what an assistant can find and what it will repeat, because a synthesis of six sources weighted toward the ones that agree is not going to be led by your own marketing copy.

The practical version is unglamorous. Write down the facts only you can state, put dates on them, and make them easy to lift. Price, limits, what changed, who does the work, what you measured. Then go and earn the third tier the slow way, by being worth referencing. There is no faster route, and the people selling one are selling the thing Google names as site reputation abuse.

One subject does not mean one artifact copied to ten places. It means the same expertise published in the shape each surface rewards, which is usually five different deliverables rather than five uploads.

This is the second rule I would put on the wall, and it is the one that saves the most wasted work. "Be on more surfaces" gets heard as "post the article everywhere", which produces five copies of one thing, none of them suited to where it landed. Adobe reaches the same conclusion from the enterprise end, recommending that one high-authority asset be adapted for multiple surfaces rather than maintaining disconnected versions of the truth.

The AI row in both of those examples is the one people find strange, so it is worth being explicit. You do not publish for AI search. You get retrieved by it, and the unit that gets retrieved is a passage rather than a page. Which means the practical requirement is that a section of your page has to make sense when it is lifted out of the page, with no back-reference to an earlier section holding it together.

Audience, then job, then surface, then format, then outcome. In that order, and never platform first.

The opponent is the way this decision is normally made, which is that somebody senior read an article on a plane. Platform-first planning produces a channel with no job attached to it, and a channel with no job attached cannot be evaluated, which is why those accounts never get closed.

Run it forwards. Who is the buyer. What is the job they are doing at the point you want to reach them. Which surface serves that job for that buyer. What format does that surface reward. And what outcome would tell you it worked. Five answers, and if you cannot fill in all five, you have a channel idea rather than a plan.

Here is a tool that does the surface step. Three questions, and it hands back a ranked shortlist plus, more usefully, a list of what to drop.

Take the bottom half of that output seriously. Where you should not spend time is as important as where you should, and it is the half nobody writes down. For a B2B SaaS company Pinterest is almost certainly irrelevant. For wedding inspiration it might be the whole business. For emergency plumbing, Maps beats every AI surface on this page combined, and I would tell that client to ignore ChatGPT entirely this year without any hesitation at all.

The tool is my judgement with numbers attached, not a calculation, and the weights are readable in the repo. Argue with it using what you know about your buyers. If you and it disagree, you win.

Measure each surface in its own report, measure the business outcome once, and accept that the line between them is a judgement rather than a calculation. There is no single number across six surfaces, and a tool that shows you one has invented it. Which is also the short answer to "what tools do I need": the free first-party report for each surface you kept, and nothing else until those are being read. The paid options, with prices and what each one is for, are in the SEO tools directory.

The opponent is the dashboard promise: one screen, one score, all channels reconciled. It cannot exist, because the surfaces do not share a unit. A YouTube search impression, a Copilot citation, a Reddit mention and a Maps direction request are four different events, and adding them together produces a number with no meaning that somebody will nonetheless put in a board deck.

So keep it light at this stage. Visibility per surface, engagement per surface, and then the business outcome, which is the only row that pays for anything.

One receipt on the AI row, because it is the row people most want a clean number for. On zplatform.ai, a property I own, ChatGPT sessions land under four different GA4 channel labels in the same report across 1 January to 22 August 2026. Reading only the AI Assistant row undercounts ChatGPT on that property by about 42%. That is not a configuration mistake, it is what channel grouping does to a referrer that arrives several ways, and the full split with the screenshot is in the measurement chapter of step two.

Then the harder problem, which is not a tooling gap and cannot be bought out of. Somebody discovers you on YouTube, asks an assistant about you, reads Reddit, Googles your name, and buys directly four days later. Which channel gets the credit? Semrush answers this honestly rather than selling a fix: attribution will be imperfect, and the value builds over time. Here is what that looks like, step by step.

You do not need to solve that at step one of a roadmap. You need to know it is there, so that when a report tells you Direct is your best channel you recognize what you are looking at.

Eight claims I have been told about this subject, all of them wrong, most of them by somebody who was selling something. Two of them are answered by the same study that supports the rest of this page, which is the test of whether I read it or quoted it.

Twelve questions. Every answer is sourced, and every one of them targets a distinction that changes a decision rather than a definition you could look up.