Can we do AI search optimization in-house?

Yes — the method is public and none of it is technically difficult. The constraint is almost never capability; it is whether someone in your company can reliably spend a few hours a month on it forever, because the work only compounds if it is sustained.

What is genuinely easy to do yourself

• The audit. Fifteen prompts across three engines, about an hour. Full method here, free.
• Technical hygiene. Checking crawler directives, adding a sitemap, writing an llms.txt. An afternoon.
• Claiming directory and association profiles. Tedious, not hard, and often the highest-return hour available.
• Getting spec tables out of PDFs. If you have someone who can edit the site, this is a day and it is worth more than most content.

What people underestimate

Sustained publishing volume

One good page does very little. The programs that move are publishing consistently — call it four substantial pages a month — for six months or more. Almost every in-house effort we have seen described starts strong, produces three pages in month one, and stops in month two when a customer escalation lands. The work is not hard; it is relentless, and it has no deadline forcing it, so it always loses to work that does.

Writing that survives an engineer's read

Industrial content fails in a specific way: it is generically correct and specifically wrong. A tolerance stated in the wrong units, a material grade that does not exist in that form, a certification described loosely. Your buyers are technical, and one visible error costs more credibility than five good pages earn. This is also the trap with using an AI writing tool unsupervised — the output is fluent, confident, and occasionally invents a specification.

Knowing what to write about

The instinct is to write about what you want to sell. The pages that get cited answer what buyers ask, which is usually more granular and less flattering — selection tradeoffs, sizing methods, failure modes, honest comparisons. Getting that list right is what the citation trace is for.

A reasonable in-house plan

1. Run the audit. Log who is named and which sources were cited.
2. Fix the technical basics in one afternoon.
3. Claim every directory and association profile in your category.
4. Pick the five questions from the audit where you were absent and your expertise is strongest.
5. Publish one page per month against that list, answer-first, with a real spec table.
6. Re-run the same audit monthly and log the delta.

Budget three to five hours a month, and put it on one named person's calendar as a recurring commitment. If you can hold that for six months, you do not need us.

The hybrid that usually works best

Keep the parts that need your relationships and judgment in-house: directory profiles, distributor coordination, and the technical review of anything before it publishes. Outsource the part that needs consistency: research, drafting, structuring, schema, and the monthly tracked-query log. That is the split our engine is built around — one short async questionnaire from your side, everything else from ours, and your team reviews before anything ships.