The first HTML has no complete sentence
Hero, loop, and a three-word claim. A retriever gets scraps. We write or mark the sentence that must exist in HTML.
AI Search Optimization
AI Search Optimization to Give Retrievers a True Sentence
AI search lifts text. A first screen that is only motion and a slogan gives it nothing complete to take. This child works extractability across AI search as a class: which surfaces pull, which sentence they should pull, whether that sentence is true, then which named child to open. It is not how models describe you. It is not one backlog for two search UIs. It is not GEO inclusion. It is not a ChatGPT retainer.
- Name the UIs
- Mark the sentence
- Prove it in HTML
- Hand the next URL
Position
If the HTML never states the answer, nothing honest can be lifted.
A person can infer meaning from a layout. A retriever cannot. Search intent on the live document has to be the question that passage answers. Keyword targeting is one job: that sentence settles one question. If the first screen is a canvas and a three-word claim, the class job is this URL.
Brand description in model answers
LLM SEO owns how assistants name you. Use this URL when a retriever has no true sentence to lift. Use the hub when the generated story is wrong even though the page already speaks.
Open LLM SEOOne backlog across two search UIs
AI SEO owns diagnose, split, ship, and report for classic results and AI search together. A passage ticket can live on that list. The extractability work stays here.
Open AI SEOInclusion into a synthesized answer
Generative engine optimization owns fetch, chunk, and include. This page owns whether the chunk is worth taking. Open GEO when the sentence exists and still never folds in.
Open generative engine optimizationA named ChatGPT or Google child
ChatGPT visibility and Google AI search optimization are platform children. This URL is the class. We route after the passage check. We do not start as a single-product pack.
Open ChatGPT visibilityWe treat AI search as a family of retrievers, not as one logo. We mark the passage in published HTML. We check it against facts lower on the page. We then route to ChatGPT, Google AI search, GEO, or the LLM hub. Entity SEO still names the thing. On-page SEO still writes headings. This child owns whether a true sentence exists to lift.
We will not sell a ChatGPT place as this product. We will not treat inclusion into a synthesis as this H1. We will not reprint the AI SEO backlog. We will not promise Google AI Overviews (AIO). If the page never states the answer in HTML, we name that before we open a platform child.
The LLM hub samples how you are described. This page asks whether a passage exists to retrieve. Those are different leaks. Keep both URLs. Do not merge the briefs.
GEO asks whether a generative system can fetch, chunk, and fold that passage. This page stops at worth retrieving. Open GEO when the sentence is already there and a synthesis still never uses it.
When this fits
When AI search would leave with fragments
Use this child when the live page answers in design, not in a sentence. If the leak is brand description, a shared backlog, or GEO inclusion, start on those URLs.
The real reply sits behind a click the fetch never makes
The question is real. The evidence is in an accordion the crawler skips. We name the extractable block. FAQ schema is not a substitute.
The same URL is rewritten once per AI product
Teams ship ChatGPT copy, Perplexity copy, and Google copy as if they were different sites. The class job is one passage. Platform children come after the check.
A vendor sold a platform place as extractability
A passage review can still ship. A purchased ChatGPT slot is not this deliverable. If the unknown is tickets across two UIs, that is AI SEO.
Process
Name the UIs, mark the sentence, prove it, hand the next URL
We work in this order. We do not open a named-assistant retainer as the first move. The point is a true sentence a retriever can lift.
-
Name the UIs
We list AI search as a class: answer modules, assistants, and other retrievers buyers already use. We do not pick a winner logo first.
Outcome A surface list for this URL, without a platform score.
-
Mark the sentence
We mark the line that should be lifted. Topical relevance is whether nearby paragraphs support it. Keyword density is not an extractability score.
Outcome A marked passage in published HTML, or a gap.
-
Prove it in HTML
We fetch the page as a retriever would. We confirm the sentence is true, complete, and not contradicted below.
Outcome A pass or fail on lift and truth.
-
Hand the next URL
We send inclusion work to GEO, description sampling to the LLM hub, ChatGPT-specific checks to that child, and Google AI search work to that child. Google AI Overviews (AIO) is named only as a surface we do not sell. Generative search notes stay observational.
Outcome A route note with the next live URL.
Deliverables
What editors can keep after we leave
You leave with a surface list, a marked passage, a truth check, and a route. You do not leave with a ChatGPT pack, an Overview, or a GEO synonym for this H1.
Surface list
AI search as a class for this URL. Dated. A vendor logo sheet is not this list.
Marked passage
The sentence in HTML that should be lifted, or the gap if none exists. Line-level, not a mood board.
Truth check
Whether the passage matches facts elsewhere on the page. Contradictions are listed.
Route note
Which child URL is next. Platform packs are not implied.
First-screen note
What a fetch sees before interaction. Mixing two answers on one hero is how a retriever lifts the wrong SKU. We list which URL owns which question. An acquired product line can keep a second URL.
Benefits
What a hero loop cannot replace
Retrievers get a complete sentence
The first HTML states the answer. Slogans stop being the only lift.
Platform rewrites wait their turn
The class passage exists first. ChatGPT and Google children route after.
False sentences are caught
The check compares the lift to the rest of the page. Invented claims are named.
The next door is written
GEO, LLM hub, or a named assistant. Nobody pretends this URL is all of them.
Methodology
Who can put a full sentence in the first HTML
This works when someone can edit the first screen, and someone can confirm the sentence is true. If both are missing, we are annotating a screenshot. We are not making passages.
We start with the live URL, the buyer question, and the first HTML. If any of those is missing, we name it first.
We fetch published HTML. We do not trust the design file. The passage has to match the question AI search will ask.
You may rename the roles below. You may not leave the lift-able sentence implied. An implied answer is how a retriever invents one.
- AI search optimization lead SEOConsultants.ai marks the passage, runs the check, and writes the route. We do not become your ChatGPT agency because the hero was empty.
- Whoever edits the first screen Can put a complete sentence in HTML. A locked hero with no editor is a finding.
- Subject owner Confirms the sentence is true. A marketing claim with no owner is a finding.
- Whoever already runs SEO Keeps the URL on one primary question. They do not mint a second site for each assistant.
- Leadership Accepts that a platform place is not for sale. An unread passage mark is not a failure of this page. A purchased ChatGPT pack would be.
First working session
Start with the first HTML, not the animation
The first hour is the fetched first screen and one buyer question. It is not a tour of AI search products.
We fetch the URL. We look for a complete sentence that answers the question. If none exists, that is the job. If the real question is how models describe you, that is the LLM hub. If the real question is one backlog, that is AI SEO. If the real question is inclusion into a synthesis, that is GEO. If the real question is Google’s AI search UI, that is the Google child after this check.
Then we compare the candidate sentence to the rest of the page. Contradictions are the job. A ChatGPT paste is not evidence of extractability.
We leave with a marked passage or a stop. A mark names the sentence and the route. A stop names what is missing: an editor for the hero, a true claim, or a request that is actually a platform pack. Both are outcomes. A tool roundup is not.
If the first hour shows the sentence exists and still never appears in a synthesized answer, we hand off to GEO. This child still owns extractability. It does not become retrieval because the passage was fine.
AI search
A lift-able passage is not an AI Overview
True sentences can help later extraction. This page cannot sell AI Overviews, a chat mention, or LLM visibility as a score.
A cleaner first screen does not buy a place in Google AI Overviews (AIO). If there is no sentence, generative search has nothing honest to lift. That is ours to tidy as extractability. If the passage already tells the truth and a module still omits you, we route to GEO or a platform child. We do not sell Overview presence from this page.
The passage has to print in the first HTML
AI search needs text it can fetch. A canvas-only hero does not help. We do not sell Overview presence.
Names in the passage should match the rest of the page
We check product and brand names in the lift-able sentence. That is entity optimization as an extractability input. Full identity specs stay on Entity SEO.
Measuring descriptions is another URL
LLM visibility needs sampled answers on the LLM hub. This page measures lift. Semantic keywords support the passage. They do not replace it.
Schema
Markup points at a passage the HTML already has.
JSON-LD can label a question and answer the page already shows. This work refuses schema as a substitute for a missing sentence. It is not a schema campaign as the whole job.
FAQ or QA markup only for visible answers
If schema has an answer the HTML does not print, retrievers get a lie. We name it. We do not stuff extra questions.
One answer block
Two plugins injecting two answers is a passage bug. We pick one speaker. Valid types still need the sentence in HTML.
Extra types do not buy a lift
Markup will not force generative search to cite you, and it will not force Google AI Overviews (AIO). Eligibility is not a purchased mention.
Who we work with
The person who can edit the first screen
Someone has to put a sentence in HTML. We work with that person. We do not replace them with a second AI writing retainer.
First-screen editor
Ships the complete sentence. Rechecks after a hero redesign.
Subject owner
Confirms the sentence is true. Holds the line when design wants only a slogan.
Whoever already runs SEO
Can keep going once the passage is marked. If they try to run this as the LLM description hub, two jobs collide. Descriptions stay on LLM SEO.
Questions
Frequently asked questions
What does this page actually change on a live URL?
It changes whether a retriever can lift a complete, true sentence from published HTML. We name the AI search surfaces that already try to pull from that URL. We mark the sentence. We check it against the rest of the page. Then we point to a named platform child if one surface still needs its own checks.
Why is this not the LLM SEO hub?
The hub asks how language models describe the brand as a story: name, mix-ups, invented products. This child asks whether a passage exists that those systems can extract. A retired SKU in an assistant reply is the hub. A hero that never prints a sentence is this URL.
Why is this not the AI SEO program?
AI SEO is one operating plan for classic Google results and AI search UIs. Tickets, owners, and reports live there. This page is the extractability craft: a lift-able sentence across AI search as a class. A passage ticket can sit on that backlog. The craft still lives here.
Why is this not generative engine optimization?
GEO traces fetch, chunk, include, and sample into a synthesized answer. This page stops when we know a passage is worth retrieving. If the sentence exists and a synthesis still never uses it, open GEO. Both URLs stay live. There is no /geo/ path.
Is this a ChatGPT package?
No. ChatGPT visibility is a named child. This URL is the class of AI search retrievers. We may route to ChatGPT after the passage check. We do not sell this H1 as one assistant product.
Does a better sentence put us in Google AI Overviews?
No. We do not sell a place in Google AI Overviews (AIO), a generative-search mention, or an LLM visibility score. A true sentence can help later extraction. Lift is what we check. Overview presence is not a deliverable.
Should the sentence match what a person would search?
Yes. The passage should settle the question a retriever will ask from that URL. Stuffing extra phrases into an overlay is not that job.
Extractability, not an assistant pack
If a retriever fetched only the first screen, would it get a complete true sentence?
Share the live URL, the buyer question you think that URL should settle, and what the first HTML actually prints. We mark a lift-able sentence or we name the gap. Then we say whether GEO, the LLM hub, or a named assistant child is the next door.
- AI search named as a class
- One sentence marked in HTML
- Truth checked against the rest of the page
- Next child URL written down
Class of retrievers first. Named products after the sentence exists.