Google names the losing kind of content.
It is what most agencies sell.
Its guidance for generative AI features says what wins: unique, non-commodity content with a point of view that could only come from experience. The example it gives of the opposite is a generic listicle of tips. That describes the majority of what gets published under the heading "content marketing" — including, until this rewrite, the page you are reading.
Built for home services contractors, franchise systems, private equity portfolios and mid-market operators — the businesses whose customers research for months before they call. One location or two hundred, the content either answers the real question or it does not.
Content used to compete for a click. Now it competes to be quoted.
A ranked page gets read by whoever clicks it. A quoted page gets read by everyone who asked the question, whether they click or not. Those are different jobs, and they reward different writing.
Google's guidance for generative AI features is unusually specific about what it rewards. It asks for unique, non-commodity content that brings something new — and its example of the failure case is a listicle of generic tips that restates what is already everywhere. Its example of the opposite is a first-hand account of a decision and what it actually cost.
That distinction is not a style preference. It is mechanical. A generated answer is assembled by retrieving pages and reviewing what is on them. Content that says what forty other pages already say gives a retrieval system no reason to choose it — and forty near-identical candidates is exactly the situation most commodity content creates.
What survives is the specific: the number you actually saw, the decision you actually made, the thing that went wrong and what it cost to fix. That material has a property nothing else on this page does — a competitor cannot copy it and a model cannot generate it, because neither of them was there.
The rest of the discipline is unchanged. Content still has to be helpful, reliable and people-first, still has to be findable, still has to convert. What changed is that the bar for "worth publishing" moved from competent to otherwise unavailable.
The same subject, twice. Only one of them is worth publishing.
Both passages below answer the same question for the same audience. The left is what most agencies ship. The right is what the same brief produces when someone who has actually done the work is in the room.
"Why maintenance agreements matter"
- Regular maintenance is essential for keeping your equipment running efficiently and avoiding costly breakdowns.
- A well-maintained system lasts longer, uses less energy, and gives you peace of mind all year round.
- Our team of experienced professionals is committed to delivering exceptional service tailored to your needs.
- What is wrong with it: every sentence is true, and not one of them could only have been written by you. Swap the logo and it is a competitor's page.
- What a model does with it: nothing. There is no fact in it worth retrieving and no claim worth attributing.
"What we learned auditing 400 lapsed agreements"
- Agreements sold at the point of a repair renewed at roughly twice the rate of agreements sold cold — and the gap held across every technician.
- The single largest cause of non-renewal was not price. It was that nobody had been to the property between the sale and the renewal notice.
- Two visits a year beat one visit plus a discount, in every cohort we looked at.
- What is right with it: every line is a finding. None of it is available anywhere else, because it came out of work someone actually did.
- What a model does with it: quotes it, and names the source, because it is the only page that says it.
Before anything is written, it has to survive three questions.
These are drawn from Google's own guidance rather than from an agency framework. A brief that fails any of them produces content that will be competent, publishable, and pointless.
Could only we write this?
Its helpful-content guidance asks whether content demonstrates first-hand expertise and depth of knowledge — whether it shows that you have actually used the product, visited the place, or done the work.
- Recorded interviews with the people doing the work, not a form sent to marketing
- Real numbers from real jobs, with the assumptions behind them stated
- The decisions that went badly, which are usually the most useful part
- If nobody can supply anything only you know, the piece does not get commissioned
Does it add, or restate?
Its AI guidance warns against creating separate content for every query variation and states plainly that a high quantity of pages does not make a site higher quality. Its spam policies define scaled content abuse as producing many pages primarily to manipulate rankings rather than to help people.
- Publishing calendars built on keyword-variation counts
- The same article rewritten for eleven adjacent phrases
- Volume commitments that outrun the supply of things worth saying
- Any brief whose only justification is that a competitor has one
Would a reader trust it?
Whether the content presents information in a way that makes a reader want to trust it — clear sourcing, evidence of expertise, and a background on the author or the site publishing it.
- Statistics cited to a blog that cited a blog that cited nobody
- Figures with no date, so a 2019 number reads as current
- Claims attributed to "studies show" with no study named
- No author, no review date, no way to tell who is accountable for the claim
The process is mostly extraction. The writing is the easy part.
Most content programs start with a keyword list. Ours starts with a calendar invitation, because the constraint is never words — it is access to the person who knows the answer.
The editorial sequence
Nothing below is exotic. It is unusual only in that most programs skip the first two steps entirely and start at drafting, which is why most programs produce the left-hand column.
- Extraction. Recorded conversations with the operators — owner, lead technician, intake, whoever holds the knowledge. Transcribed, mined for the specifics nobody thought were interesting.
- Claim inventory. Every factual assertion the piece will make, listed before drafting, each one assigned a source: your data, a primary source, or an observation labeled as one.
- Sourcing pass. Each external claim is checked against the issuing authority — the body that actually publishes it — with the exact wording read, not the topic assumed.
- Drafting. Answer-first structure so the reader is satisfied in the first paragraph and a retrieval system can lift a coherent passage without distorting it.
- Editorial review. Read cold against the claim inventory. Anything that cannot be sourced is cut rather than softened into vagueness.
- Operator sign-off. The person interviewed confirms the specifics are right before it publishes. They usually catch one thing, and it is usually the number.
Six assets that do different jobs — and what each one is actually for.
Not a content menu. Each of these exists because it does something the others cannot, and we build the ones a business actually needs rather than all six by default.
Decision pages
The page someone reads while deciding whether to call you — comparing two options, working out what a job involves, checking whether their situation is the normal one.
Why it exists: it is the only content type that catches people at the moment of intent, and it is the one most sites do not have because it is harder to write than a blog post.
Question-level answers
Direct answers placed where the question is actually asked, written to satisfy a reader in a paragraph and to be liftable without losing meaning.
Why it exists: a generated answer is assembled from passages, not pages. Content structured as one long argument with the payoff at the end gets skipped for content that answers immediately.
Operational deep dives
The long-form piece that only exists because someone did the work and counted the results. Findings, not tips.
Why it exists: this is the asset that earns citation, links and the reputation that makes everything else easier. It is also the one that most often does not get commissioned, because it needs a real interview.
Service and category pages
The commercial pages, written to convert rather than to fill a keyword slot — with the objections handled where they arise instead of in a FAQ at the bottom.
Why it exists: these carry the revenue. Treating them as SEO furniture while the blog gets the attention is the most common misallocation we inherit.
Refreshes and consolidations
Existing content either updated with something new or merged into the piece that deserves to survive. Google picks one canonical per duplicate cluster regardless of what you would have chosen.
Why it exists: most sites do not have a content shortage. They have twelve thin pieces on one topic where one good piece would rank, get cited and convert.
The technical layer
Titles and descriptions written rather than generated, Article structured data where it fits, internal links that follow priority instead of chronology.
Where we stop: Google states structured data is not required for generative AI search and no special markup exists for it. We deploy it for what it does do, and bill it as that.
We rewrote this page for the reason we would rewrite yours.
We run content programs for single-location contractors and for national franchise systems and private equity portfolios. The volume changes enormously between them; the standard does not, because a homeowner reading a page cannot tell how large the company behind it is and does not care.
The previous version of this page carried thirteen outbound links. Every one of them pointed at a directory listing for Allegiant. Not a single citation, not a single source, and no mention anywhere of AI, citation, or first-hand experience.
It was competent, it was on-brand, and it was the exact commodity content Google names as the failure case. An agency that cannot fix that on its own site has no business charging to fix it on yours.
So the standard is visible rather than promised. Every claim here links to the documentation behind it. Where we describe something no authority has published — how ChatGPT or Perplexity choose what to quote — we say that it is observation. That is a harder page to write and a much easier agency to check.
What separates this from a standard content retainer.
Not a competitor teardown — a description of where the practices differ, so the choice can be made on substance.
| Practice | Standard content retainer | AllegiantOMNIVIZ™ |
|---|---|---|
| Where a brief starts | A keyword list | An interview with someone who did the work |
| Volume | Fixed monthly count | However many things are genuinely worth saying |
| Sourcing | Whatever ranks for the stat | Issuing authority, exact wording read |
| Thin existing content | Left in place, more added | Consolidated into the piece that deserves to survive |
| Query variations | One page each | One page, because Google says the split does not help |
| Author accountability | Byline optional or generic | Named author, review date, operator sign-off |
| Reporting basis | Traffic and rankings | Measured data separated from observed sampling |
Four content line items you can stop paying for.
Each is drawn from Google's own mythbusting guidance. If a content proposal on your desk includes one of these, read the source before signing.
Rewriting content into an "AI-friendly" voice. Not needed. The systems understand synonyms and general meaning — which is also why generating a separate page for every phrasing of the same question is wasted budget and a scaled-content risk.
Chunking articles into fragments. There is no requirement to break content into small pieces for AI to understand it, and no ideal page length exists. Structure for readers.
AI text files and special markup. Google states it does not use AI text files such as llms.txt, and that structured data is not required for generative AI search. Article markup is worth deploying for what it actually does, which is a different and honest reason.
Volume commitments detached from substance. A high quantity of pages does not make a site higher quality, and mass-producing them to manipulate rankings is named in the spam policies. Twelve pieces a month is only a good number if there are twelve things worth saying.
The pattern beneath all four: there is no format trick. There is content someone with real knowledge stood behind, and content there was no reason to publish.
- Google Search Central — Creating helpful, reliable, people-first content
- Google Search Central — Optimizing for generative AI features
- Google Search Central — AI features and your website
- Google Search Central — Spam policies for Google web search
- Google Search Central — Article structured data
- Google Search Central — Control your snippets in Search
- Google Search Central — Influencing your title links
- Google Search Central — Consolidate duplicate URLs
- Google Search Central — SEO starter guide
- Google Search Central — Evaluating third-party SEO advice
- Search Console Help — Generative AI performance report
The questions operators actually ask.
Answered against primary sources where they exist, and answered honestly where they do not.
How much content should we publish each month?
Fewer pieces than most retainers promise. Google's AI guidance states directly that a high quantity of pages does not make a site higher quality, and warns against creating separate content for every query variation. The honest number is however many things you have that are worth saying, which for most businesses is two or three substantial pieces rather than twelve thin ones. If a proposal quotes a volume before anyone has asked what you know, it is selling typing. Consolidation of what already exists usually beats addition — that work sits alongside technical SEO.
Can we just use AI to write it?
For the commodity half, it makes no difference — and that is the problem. Google's position is about quality rather than production method, but its helpful-content guidance asks whether the content demonstrates first-hand expertise, and generated text cannot demonstrate what nobody did. A model can draft, structure and tighten. It cannot supply the number you saw on a job last March. We use AI where it helps and interview humans for the part that carries the value, which is the same division described on our AI SEO practice page.
What actually makes an AI system quote us instead of a competitor?
Being the only page that says the thing. Google documents that its generative features retrieve relevant pages and review the information on them to build a response, and that what performs is unique, non-commodity content with a point of view. Where forty pages say the same thing, there is no reason to pick yours. Where one page carries a specific finding, there is nowhere else to get it. Beyond Google's own surfaces this is observation rather than documentation — ChatGPT, Perplexity and Claude publish no equivalent account — and we label it that way. Measurement runs through AI visibility monitoring.
Do we need to restructure our content for AI?
No. Google states there are no additional requirements or special optimizations to appear in AI Overviews or AI Mode, no requirement to chunk content, and no ideal page length. What does help is ordinary good structure — answering the question near the top rather than after eight hundred words of preamble — because that serves a reader first and happens to make a passage liftable. Any proposal that prices "AI restructuring" as a separate service is pricing something the issuing authority says is unnecessary. Related: content writing.
We have 200 old blog posts. Delete, update, or leave them?
Sort them into three piles. Posts with a genuine evolution story get updated — what was true then, what changed, what you would do now — and become internal-link fuel. Posts that merely restate what everything else says get consolidated, since Google selects one canonical per duplicate cluster and the rest spend crawl budget to be ignored. Posts that demonstrate you stopped paying attention — a 2023 SEO piece with no mention of AI — actively undermine you and should go. See the blog for what the surviving set looks like.
Who writes it — you, or us?
We write it. You supply the part nobody else can: forty-five minutes of recorded conversation per piece with whoever holds the knowledge. That is the entire ask, and it is also the thing that determines whether the piece passes Google's first test — whether it demonstrates first-hand expertise. Programs that skip the interview to save the partner time produce content the partner then does not want to publish, which is a more expensive way to save an hour. Operator sign-off before publication is part of the same loop, alongside market research.
How do we know whether it is working?
In two layers kept separate. The measured layer is Search Console, including its generative AI performance report, which covers impressions from AI Overviews and AI Mode. The observed layer is sampling the assistants on a fixed question set and recording who gets named. The second is genuine signal and is not a ranking metric — Google states no third-party tool has access to its internal ranking or AI systems, ours included. Conversion is reported against the pages that actually produced calls, which pairs with conversion rate optimization.
Does content marketing still pay when nobody clicks through?
The honest answer is that the mechanism changed and the accounting has to change with it. Being quoted in an answer reaches everyone who asked the question, whether or not they click, so click-through alone now understates the return. It also overstates it in the other direction, because a brand named in an answer without a link earns awareness that no analytics package will attribute. We report what is measurable, label what is observed, and do not invent a number for the gap between them. Google's own guidance on evaluating third-party SEO claims is a useful lens on anyone who does. Related: competitor analysis.
Should our titles and descriptions be written or generated?
Written. Google explains that it generates title links from page content and may use something other than your tag when the tag is poor, and it documents how snippets are formed. Writing them well is the cheapest content work available: it changes what a person sees before deciding to click, on pages that already exist. Every page we build ships with both written to a fixed standard rather than auto-filled, and mirrored into the page's structured data so the two cannot drift apart. Related: landing pages.
How long before content produces anything?
Consolidation and rewrites of existing pages move fastest, because the pages are already indexed and the change is to what they say. New work compounds over months, because citation and linking accumulate rather than switch on. Nobody should give you a date: Google states that indexing and serving are not guaranteed even when everything is done correctly. What we commit to is a claim inventory per piece, sources you can check, and reporting that separates measured from observed. If you want the current picture before deciding, start with an A.R.C. Report.
Find out what your content is actually saying.
A content audit: which pages carry something only you could have written, which restate what everyone else already published, and which should be consolidated rather than kept. Findings are yours whether or not we work together.
- Every page classified as unique, commodity, or consolidation candidate
- Claim audit — which assertions are sourced, which are unsupported
- Duplicate cluster map with the canonical you would have chosen
- Retrieval test on a fixed question set across the major assistants
- Findings ordered by consequence, with the evidence tier on each
Explore the wider program: all services, local SEO, SEO, answer engine optimization, reputation management and email marketing.
We will show you which of your pages carry something only you could have written — and which ones a competitor could have published word for word.
No cost, no commitment. We will follow up by email or phone to walk you through the findings.
