There is a real frustration moving through SEO, publishing, and marketing. For years the instruction was simple. Create helpful content. Answer the question. Serve the user. Earn the ranking.
Then the answer layer arrived. A search engine now summarizes the work before the user clicks. An assistant blends several sources into one response. A potential customer gets the definition, the comparison, the recommendation, without ever visiting the page that made the answer possible.
So when Google tells site owners to create more unique, non-commodity content, many people hear something else: produce better material for the machine to absorb.
That reaction is fair. The old search bargain was not perfect, but it was legible. You published, Google ranked, users clicked, and some of that attention became leads, sales, or brand memory. AI search changes the shape of the deal. The answer can now happen before the click.
That does not make content useless. It makes replaceable content nearly defenseless.
And the data is not theoretical. In the first four months of 2026, about 68 percent of US Google searches ended without a click, up from roughly 60 percent in 2024 (SparkToro, using Similarweb clickstream data). Tracking real browsing behavior, Pew Research found that users clicked a traditional result only 8 percent of the time when an AI summary appeared, compared with 15 percent when it did not, and clicks on the sources cited inside those summaries sat at about 1 percent.
That does not mean rankings stopped mattering, or that every industry loses equally. It means the old assumption is breaking. Your content may be retrieved, summarized, even cited, and the user may still never arrive.
Notice what is quietly being asked of you here. Not a tool. A partnership. The work now wants a human who can produce what a machine cannot, and a machine that can carry it where a human cannot reach. Nobody requested this arrangement. It was not pitched in a strategy deck or voted on by the people it reshapes. It arrived as a side effect of someone else’s product decision, and then it became the job.
One definition up front, because the title turns on it. By a real copywriter, this essay does not mean someone who decorates a keyword brief. It means the person who interviews the business, finds the proof, challenges the vague claim, and turns what is true into language a buyer can trust. Hold that distinction; the rest of the piece depends on it.
When the page could have been anyone’s
Replaceable content is content almost anyone could have published. It may be accurate. It may be well written. It may even rank. But it does not contain enough ownership, proof, or specificity to belong to one business.
A “what is SEO” article is usually replaceable. So is “benefits of HVAC maintenance,” “how to choose a personal injury lawyer,” and “five tips for better digital marketing.” The topics are not bad. The problem is that most versions of those pages are interchangeable. They restate public knowledge in slightly different language, which is exactly what an AI system is built to summarize, blend, and replace, because nothing inside is something only that business can credibly say.
That kind of page answers a question without creating preference. It informs without proving. It fills space without building an asset.
In traditional SEO you could often win anyway, by publishing a decent answer to a common question and optimizing it well. In AI search that same page is still useful to the system and less useful as a destination. The content becomes input. The business may not get the click.
Google is right. Publishers are also right.
Google’s advice to create unique, non-commodity content is not wrong. In its first official guidance on the subject (May 2026), Google explained that its generative features run on the same core Search systems, using retrieval and query fan-out to surface content from its index. It also said, plainly, that you do not need an llms.txt file, special schema, AI-specific markup, or tiny AI-friendly chunks to appear.
A lot of the market wanted a shortcut. A file. A tag. A schema trick. A formatting rule. The official guidance points somewhere less convenient: a clear point of view, content that goes beyond common knowledge, helpful and people-first material, real images and video, and a clean technical structure.
The advice is technically reasonable. The frustration is economic. The same environment asking for better content can summarize that content before the user clicks. Both things are true at once. Google can be right that generic content is weak, and publishers can be right that the exchange has become less fair. That tension is the new normal, and pretending either side is lying does not help anyone manage it.
There is a third party in this, usually offstage. The companies that built the answer layer are not working from a finished plan either. You can watch them improvise in real time: publishing official guidance one quarter, clarifying it the next, telling the market to ignore the very files and tactics a whole cottage industry sprang up to sell. That is not the posture of an institution that fully modeled the consequences before shipping. It is the posture of one learning them in public, while an open web it quietly depended on gets rerouted through its own front page. The curtain that let platforms describe themselves as neutral plumbing is thinner now, because the disruption has names, livelihoods, and a measurable click count attached to it. Big tech is in a new world too. It just has better lawyers.
Proof can still be summarized. That is not the point.
Here is the trap in most of this conversation, including the version many SEOs tell themselves. People assume proof-rich content is safe because AI cannot summarize it. That is not true. An AI can compress “what a firm sees in the first 30 days of a case” as easily as it compresses a generic explainer. Specificity does not make content impossible to summarize.
What specificity does is make the content harder to fake and more valuable. It makes the business the obvious source, the named example, the thing the reader remembers and returns to. A generic page, once summarized, costs you nothing, because it gave away nothing that was yours. A proof-rich page, once summarized, still works for you, because the proof points back to a business that can be selected, trusted, and named.
That distinction matters, because being in the answer is not the finish line. A page can be cited and barely shape the response. A brand can be mentioned and not recommended. A company can appear in an AI answer while the framing quietly favors a competitor. The gap is real: in large retrieval studies, many of the pages an AI surfaces while assembling an answer never make it into the final response at all. Being found is not being cited, and being cited is not being chosen. The space between those rungs is where most “we got cited” strategies quietly fail.
So the better question is not “how do we get quoted.” It is this: what does the market understand about us, and does that understanding help the customer choose correctly?
Which is why copywriting matters more, not less
The common misread is that copywriting loses value because AI can generate words fast. The value of a real copywriter was never word production. AI can produce polished paragraphs all day. A copywriter is valuable because they find what is not replaceable. AI did not steal the pen. It made the empty page more obvious.
They know how to interview a business. They listen for the detail the owner skipped because it felt obvious. They turn process, proof, customer pain, objections, and lived experience into language that gives the business a sharper position. Generic content is easy to flatten. Specific content is hard to replace. The copywriter helps a company say something only that company can credibly say. That is not decoration. It is a visibility asset.
It helps to see where the role came from. For most of the last decade, the SEO content writer was largely a volume function: take a keyword brief, produce a competent answer to a common question, publish, repeat. That is precisely the job AI now does at scale and at near-zero cost. What does not transfer to a machine is the part that was always the real craft, the interviewing, the judgment, the extraction of proof. The writer who only filled briefs is being automated. The writer who uncovers what only one business can say is becoming more valuable, not less.
The difference is not abstract. Take a roofing company’s service page, the kind every contractor publishes:
Before (replaceable): Our experienced team is committed to providing quality roofing services you can trust.
After (defensible): Most of our emergency calls between November and February are ice-dam leaks on north-facing roofs built before 2005. Here is how we tell, on the first visit, whether you need a full tear-off or a targeted repair.
Both sentences are about the same company. Only one of them could not have been written by anyone else, and only one gives a homeowner a reason to remember who said it.
There is a second reason this matters. AI writing tools pull language toward the middle. They are trained to produce plausible, fluent, statistically likely text, and a Cornell study of AI-assisted writing found that AI suggestions made writers’ output measurably more similar to one another. In practice, AI-assisted copy gets smoother, safer, and less distinctive unless a human actively pulls it back toward specificity.
The market is already full of interchangeable claims. We care about our customers. We provide quality service. We are committed to excellence. None of those lines are false. Most of them are weak. Committed to excellence is not a position; it is a decorative pillow. They reveal nothing about how the business thinks and give a buyer no reason to remember the company. AI can generate that language endlessly. A real copywriter should be allergic to it. The goal is not to sound polished. It is to sound true, specific, and difficult to confuse with anyone else.
AI assists. The human owns the meaning.
None of this means copywriters should avoid AI. Used well, AI helps a copywriter move faster. It can organize notes, summarize interviews, test angles, generate alternate headlines, compare tones, and pressure-test whether a page is clear. It is useful in the workshop.
What it should not do is become the source of the company’s meaning. That is the line. AI can shape raw material. The copywriter decides what is true, what is overclaimed, what sounds generic, what needs proof, and what the business can actually stand behind. Using AI well is not outsourcing judgment. It is using a tool to sharpen it.
The danger is not that AI writes sentences. The danger is letting AI invent the substance, flatten the voice, and produce something professional that could belong to any company in the category. That is how businesses end up with replaceable content at scale. The human owns the claim. The human owns the decision about what the business is allowed to say. That is why AI does not eliminate the copywriter. It raises the bar for what a copywriter has to be.
The copywriter and the visibility architect
A copywriter does not replace search strategy. The roles are different, and the second one deserves a name.
The copywriter uncovers and expresses what makes a business specific: the proof, the customer language, the process detail, the objections, the point of view. The job is to make the business harder to confuse with everyone else.
That second role is not new so much as renamed by necessity. The SEO professional has already moved through several versions of the job: the keyword optimizer who matched pages to queries, then the technical specialist who owned crawlability, site speed, and structured data. The version the answer layer now demands is broader. It is closer to an architect of how an entity is understood, retrieved, and represented across systems that no longer simply rank pages but assemble answers. Calling it a visibility architect is less about inventing a title than naming where the discipline already went.
Keep the hierarchy straight, though. The copywriter creates the meaning. The visibility architect makes sure the market and the machines can carry it. The first job is the harder, more human one; the second exists to protect it.
The visibility architect owns whether that meaning survives in the search environment. That means discoverability, indexation, content architecture, entity clarity, internal linking, schema where it earns its place, and measurement of whether the business is being retrieved, understood, cited, represented accurately, and selected. It increasingly also means earned authority off the page, because the citations now lean there. A Muck Rack analysis of more than 25 million AI citations found that roughly 84 percent came from earned editorial coverage in third-party publications, not from brand-owned pages and not from paid placements. A business’s own pages still matter, but for citation they are a minority of the signal. Being talked about elsewhere is now part of being chosen.
This is not classic SEO with a new label. Google describes its AI features as using query fan-out, generating related sub-queries and pulling from a wider pool of pages. By one large analysis, only about 38 percent of AI Overview citations also ranked in the top 10 for the same query, down from 76 percent a year earlier, though part of that shift reflects better detection in the tooling, so it is best read as direction rather than law. The point holds: ranking for one head term is no longer the only gate.
The two roles need each other. If the copy is generic, the visibility system has weak material to carry, and nothing worth earning coverage for. If the visibility system is weak, even strong copy may never be found, understood, or carried correctly into an answer. One role protects the meaning. The other protects the path that meaning has to travel.
They also fail together, and it is worth naming how. Run badly, the architect wraps clean structure and tidy schema around copy that says nothing, and the system perfectly understands a page that gives a buyer no reason to care. Or the writer produces something distinctive that no system can parse, cite, or connect, and the best sentences never reach a reader. The pairing is not two people in a room. It is two disciplines holding one standard: specific enough to matter, structured enough to survive.
Run well, though, the tandem is not only defense, and this is the part nobody put in the brochure. A human who knows exactly what is true about a business, paired with a machine that can draft, restructure, test, and scale that truth, can produce work neither could manage alone. The expert stops staring at a blank page. The model stops inventing. Ten sharp comparison pages that used to take a quarter take a week, each one carrying real proof instead of filler. A genuinely new capability is hiding inside a genuinely disruptive one, which is usually how the interesting shifts arrive, uninvited and double-edged. The businesses that feel only the threat will miss the half of this that is an opportunity.
Why this lands on the top floor
This reaches leadership faster than most marketing shifts do, and it should reach them sooner than it usually does. Organic search traffic was quietly baked into growth forecasts, which makes its decoupling from value a CFO problem, not just a content one. Traffic was the number that justified the whole function; when clicks fall while rankings hold, the case that funded the work starts to wobble even though the work is not worse, and analysts are already forecasting steep declines in classic search volume. The honest move is to change the scorecard before the budget conversation forces it: branded search, direct traffic, assisted conversions, and share of AI answers, not sessions alone.
It also moves where the money sits. A volume-content team and a proof-and-authority function are different cost structures, weighted toward fewer senior people and toward earned coverage that used to live in the PR line. Senior judgment matters more than production capacity now, because production is the cheap part. And there is a strategic edge buried in it. Because AI answers lean on earned authority and accumulated proof, a company’s reputation beyond its own website becomes a compounding asset, the kind that rewards whoever starts early and punishes whoever waits. Competitors who pair the human and the machine well get harder to displace in the answer layer every quarter. That is not a marketing tactic. It is a moat, and moats are a C-suite conversation.
And what to do now
Before publishing, the old question was: can this rank? It still matters, but it is no longer enough. Three better questions:
- 1. Could a hundred other websites publish basically the same thing? If yes, the page is probably replaceable.
- 2. Does this contain something from our actual work, customers, data, market, process, or judgment? If no, the page may be useful but it is not defensible.
- 3. After someone reads it, is there a clearer reason to trust us, remember us, or take the next step? If no, it is informational, not a business asset.
Organizations do not need to panic. They need to stop treating content as a volume game. Start by identifying the most replaceable pages: thin service pages, vague industry explainers, duplicated location pages, keyword-built articles with no business insight. Then go find the missing proof. What does the sales team explain over and over? What do customers ask before they buy? What comparisons does the business need to win? What outcomes can be shown without overclaiming? What local or category context does the business understand better than a generic writer?
Some businesses will say their best proof cannot be published, and often they are right. A law firm cannot name the client. A clinic cannot describe the patient. A vendor signed an NDA. But proof does not have to be a named outcome. It can be the pattern across a hundred cases, the aggregate number, the specificity of the process rather than the identity of the customer. “What we see in the first thirty days” is a real claim even when whose case it was stays private. Anonymized specificity still cannot be written by anyone else.
Then decide who owns which layer. The copywriter uncovers the meaning, proof, voice, and decision logic. The visibility architect structures the site, connects the content, supports entity clarity, earns outside authority, and measures whether the business is retrieved, cited, represented, and selected. AI assists both. It should own neither. It can draft, outline, summarize, and analyze. It should not decide what the business means, invent proof, or turn the company into a smoother version of everyone else.
If you are a small business, this still applies, it just lives in fewer hands. The two functions do not require two hires. They require that both jobs get done: by one capable generalist, by an owner working with an outside partner, or by a small team that knows which hat it is wearing at any moment. The resilient move for a solo operator is not to staff a department. It is to sequence. Fix the most replaceable page first, find one piece of real proof, get it structured so it can be found, then repeat. Small does not mean exposed. Undifferentiated means exposed.
The way forward, then, is not to feed the machine more generic articles. It is to build around proof, specificity, and owned value: original observation, real examples, expert interpretation, process detail, tools, comparison pages, clear entity information, and better reasons to trust the company behind the page. That is what non-commodity content actually means. Not content made because Google asked for it. Content the business needs because the answer now often happens before the click.
So where does this leave the two roles?
Not as a branded team with a clever name. The moment you package them as a product, you have rejoined the buzzword cycle this whole shift is a reaction to. They are simpler than that, and harder to copy. One owns the meaning. The other owns the path it travels.
You cannot forecast where the answer layer goes next. Citations shift month to month, a single model update can swap out who gets named, and the analyst predictions get walked back as often as they land. So you do not build for a fixed future. You build for what holds underneath every version of it. The surface is volatile and the asset is patient, and that is not a contradiction. It is the reason. Because the surface keeps moving, the durable move is to be the most credible, specific source it can find, in whatever shape it asks for next. Resilience is the achievable form of certainty.
And no, this does not commoditize when everyone does it. Generic content is interchangeable, which is exactly why it loses. Proof is not. Your thirty days of a case are not someone else’s. Your install data, your failure rates, your hard-won judgment cannot be published by a competitor, because they were never theirs to publish. Push the whole market to compete on real specificity and the contest simply moves up to who actually has the better evidence. That is a market that rewards being good, not being loud. Proof does not get used up. It compounds.
It is at least worth noticing where this points. For two decades, the open web rewarded volume over substance, gaming over honesty, and the page that was cheapest to mass-produce rather than the one that was truest. The answer layer was not built to fix that. But it may still be inverting the incentive. If the market is pushed to compete on proof, then the work gets nudged toward honesty almost in spite of itself. Nobody asked for that either. It may be the most interesting thing here.
Which brings it back to the only scoreboard that ever mattered. Not pages. Not rankings. Not even citations. You do not measure this team by what it publishes; you measure it by what the business feels. The one who owns the meaning and the one who owns the path do their work in the shadows, and that is exactly where it belongs, because the customer should never have to see the architecture. They should only arrive already feeling good about trusting you, before the first real conversation, because the proof did the persuading while they were still deciding.
That is the purpose of the whole arrangement, and it never shows up on a dashboard. It shows up when the phone rings.
TRY IT THIS WEEK
Do not start with a content calendar.
Start with the page AI can replace most easily. Open analytics, pick the single informational page that earns the most impressions, and run it through the three questions:
- 1. Could a hundred other sites publish basically this?
- 2. Does it contain proof only you could provide?
- 3. After reading it, is there a clearer reason to choose you?
If it fails the first and passes neither of the others, you have found where the answer layer is quietly taking your value. That page is where you start.
• • •
A. L. MacFarland writes on AI-mediated search, entity authority, and the economics of attention from rural farm country, Pennsylvania. This essay argues that the businesses worth saving were never the ones publishing commodity answers, and that the human-machine pairing now deciding who gets chosen was built by neither of the two people doing the work.
SOURCES
• SparkToro / Similarweb, zero-click search study, June 2026. 68.01% of US Google searches ended without a click in Jan-Apr 2026, up from 60.45% in 2024 (panels differ across years). • Pew Research Center, AI summaries and clicks, July 2025. 8% click rate with an AI summary vs 15% without; 1% clicked a cited source. • Ahrefs, AI Overview citation overlap, March 2026 (863K keywords). 38% of citations rank top 10, down from 76% (July 2025); improved parsing is a partial factor. BrightEdge reports a lower overlap (~17%). • Ahrefs, 75,000-brand study. Branded web mentions correlate more strongly with AI Overview visibility than backlinks. • Muck Rack, “What Is AI Reading?” (May 2026), 25M+ links across ChatGPT, Claude, Gemini. ~84% of AI citations from earned media; ~0.3% paid. • Google Search Central, “Optimizing your website for generative AI features,” May 2026. Core Search systems, RAG, query fan-out; no llms.txt / special markup / chunking / schema required. • Agarwal, Naaman & Vashistha (Cornell), “AI Suggestions Homogenize Writing Toward Western Styles,” CHI 2025. AI suggestions increased similarity between writers. • Gartner, forecast (February 2024): traditional search engine volume projected to fall ~25% by 2026 as queries shift to AI answer engines. A prediction, not a measured outcome.