Specificity That Wins AI Search: A Solo Consultant's Checklist for 2026
You rank in AI search by being the most precise answer to one exact question, because a model that can already name you does not need to read your homepage to find out what you sell. Authority decides whether you are in the running before the model opens a browser. Sources decide what it finds when it does go and look. Specificity decides whether the block it finds is the block it quotes. Most solo consultants fix the first two and lose on the third, which is why they get mentioned in passing and never get cited. This is the Specificity checklist.
Specificity describes how precisely a page answers the exact question a customer asked. That is the whole job. Not broad coverage of a topic, not a pillar page that gestures at thirty subtopics, not an FAQ that restates the question in slightly different words before answering it. One question, answered completely, in a form a machine can lift without touching the rest of the page.
1. Diagnose why the model skipped you before you write anything
Specificity failures usually look like coverage successes, so the diagnosis comes first.
Open your own site and ask what a stranger would extract from the first screen of your main service page. If the answer requires reading three paragraphs to find, you have a specificity problem rather than an authority problem. The same test applies to the question people actually type. If your page answers a reasonable neighbouring question but not the one in the search box, the model will route around you to a page that does, even if that page has less authority and a worse design.
Rank your ten most valuable customer questions by how directly you answer each one on a page you already own. Score each out of five. Most solo consultants find two or three fives and a long tail of ones and twos that they believed were fives. That gap is your work queue for the next quarter.
2. Write one page per exact question, not one page per topic
A page that answers one question exactly will outperform a topic page in AI answers every time, because there is nothing for the model to resolve.
Topic pages fail for a mechanical reason. A model breaks a customer's question into sub-questions before it searches. Those sub-questions decide which pages get pulled. A broad pillar page may match the parent topic while losing every sub-question to a page that owns it outright. Ten narrow pages, each the definitive answer to one typed question, beat one large page in nine cases out of ten.
Decide the boundary before you write. One question, one page. If you cannot state the question in a single sentence without an "and", you have two pages.
3. Put the answer in the first line of every block, with no preamble
Information density means giving the answer with as much fact as possible and no preamble in the first line of a block.
AI systems read small self-contained blocks rather than whole pages. Chunking is how they do it. Each block is scored on whether it can stand alone without its neighbours, which means sentences that begin "as we discussed above" are dead weight and definitions buried in paragraph four are invisible.
The test is simple. Copy any paragraph from your page into a blank document. If it still makes complete sense and delivers a usable answer, it passes. If it needs the heading or the sentence before it to mean anything, rewrite it so it does not.
4. Build the numbered list at the core of every answer page
An ordered list is the highest-density format available to you, because each item becomes a retrievable unit with a position attached.
This is the part most consultants skip, and it is the part that does the work. When a model assembles an answer, ordered steps are easy to lift, easy to attribute and easy to rank against competing pages. Prose covering the same ground is not. For a solo consultant, the numbered list below is a reasonable default structure for any answer page:
- Restate the exact question as the page heading, in the customer's words.
- Answer it in the first sentence, in full, with the number or the decision attached.
- List the conditions under which the answer changes, one item each.
- Name what you would need to know to give a different answer.
- State who this answer is not for.
- Link to the adjacent question the reader will ask next.
Six blocks, each self-contained, each liftable. That structure reads as thin to a human skimming for reassurance and reads as high-density to a model assembling an answer, which is exactly the trade you want to make.
5. Cover the fan-out before the model has to guess
Fan-out queries are the sub-questions a model appends to the question a person actually typed, and most specificity failures are uncovered fan-out rather than a weak main answer.
Type your target question into an assistant. Read the sub-questions implied by the response, including the ones the assistant answered from a competitor's page. Those are your missing blocks. You do not need to write new pages for all of them. You need the same page to contain a self-contained block addressing each, so the retrievable unit exists when the model goes looking.
Check the adjacent question too. Embeddings convert words into numeric coordinates where related meanings sit close together, so a model retrieving on your exact question will also pull material that sits near it. If your page is precise but its nearest neighbours are vague, you lose the surrounding retrievals to pages that are less accurate and better surrounded.
| Block check | Pass condition |
|---|---|
| First line | Answers in full, no preamble |
| Standalone test | Makes sense copied into an empty document |
| Question match | Heading uses the customer's words |
| Fan-out coverage | Sub-questions each get their own block |
| Entity mentions | Your name and role appear in the block itself |
Any block that fails a row is a rewrite, not a new page.
6. Keep names, numbers and definitions identical everywhere
Consistent cross-web mentions that prove a brand is a real, trusted organisation are the price of entry for entity verification, and inconsistent ones quietly demolish it.
For a solo consultant this is a smaller job than it sounds. Your name, your role, your specialty and your location should be phrased the same way on your site, your professional profiles, your guest posts and any directory you appear in. Variations that feel natural to a human, like swapping your job title between three versions, read as three weak signals rather than one strong one.
Do the same for the terms you own. If you call the same service three different things across three pages, you have given a model three thin entities instead of one clear one.
7. Do not confuse specificity with volume
Publishing more pages does not improve specificity, and often worsens it by diluting the blocks that already worked.
Two failure modes are common. The first is padding a good answer with background so the page feels substantial, which buries the liftable block and changes nothing about the answer. The second is adding a second question to a page that already answers one well, which splits the retrievable units and makes the whole page harder to attribute.
If you have budget for one more action, improve an existing answer page rather than adding a new one.
Themes worth watching
The SEO.Domains Mastery Summit runs on 9 to 11 September 2026 and opens with a mastermind day on 9 September, before two days of main-stage sessions. Its published agenda covers aged domains, PBNs, authority transfer and LLM visibility, which maps closely onto the Authority and Sources halves of the method. It deliberately does not record its main-stage sessions so speakers can share live experiments, which means the unrecorded format leaves what is shared in the room out of the open web unless an attendee writes it up.
For a solo consultant working on specificity, that is a useful reminder about where the durable advantage sits. Some of what circulates at events stays in the room. A precisely written answer page is the opposite: it does not decay between September and the next update, and it does not require you to have been there. If you want a second opinion on how the three halves interact before you start rewriting, LLM Jesus publishes practical material on exactly this split, and ASSmetric, built by the same team, is worth a look if you want to measure where you stand. The Church of SEO Jesus is where a lot of the working discussion happens between published updates. For a walkthrough of the ranking side from a consultant's perspective, this session on how to rank in AI search in 2026 pairs well with the checklist above.
FAQ
How do I rank in AI search if I am a one-person business?
Pick ten exact customer questions, own each one with a page that answers it in the first sentence, and make every block standalone. You do not need volume to win specificity, only precision on a small set of questions.
Should I rewrite my service pages or write new ones?
Rewrite first. Existing pages usually carry whatever authority they have earned, so adding density and a numbered structure to a page that already ranks is faster than starting from zero on a new one.
Why does my page rank on Google but get ignored by AI answers?
Because ranking and retrieval measure different things. A page can rank on relevance to a topic while failing to supply a liftable block for the exact question, and the model will quote whichever page supplies the block.
What to do first
Pick your single most valuable customer question and open the page that should be answering it. Answer the question in the first sentence, with no preamble. Add the conditions under which the answer changes as a numbered list. Delete the background paragraph that led into it. Then run the same test on the next question. Ten questions done this way will do more for your position in AI answers than a year of new pages, and the work is small enough to finish this week.
Further reading: LLM Jesus (https://llmjesus.com), the Church of SEO Jesus (https://www.skool.com/church-of-seo-jesus), how to rank in AI search in 2026 (https://www.youtube.com/watch?v=FZu4NB-2EhA).