// pillar
Search and AI Visibility
Two problems arrive as the same sentence: we are not getting found. In one, engines cannot index the site at all. In the other, the site ranks perfectly well and an assistant recommends a competitor anyway. Nothing else about them matches. Not the cause, not the fix, not the number that shows it changed.
This hub owns the part that decides whether you get found at all. It does not cover anything after the click. Turning attention into pipeline is the growth hub, and anything we ship as code is the build hub.
Four services sit under it, bought for four different reasons. Technical SEO decides whether engines can crawl, index and understand your own site, measured in position. AEO decides whether your words end up inside a generated answer from ChatGPT, Perplexity, Copilot or an AI Overview, measured in citation share. Link building decides how much off-site authority your domain carries, measured in referring domains earned. Outreach is the human work behind placements, measured in placements per hundred contacts.
Why these are four pages and not one
Because they fail independently. A site can be technically immaculate and still sit under three older domains. A domain with a strong link profile can be absent from generated answers because a WAF rule refuses one crawler token. Content that ranks can be unquotable, and content that gets quoted everywhere can be invisible in the ten blue links. Each failure has its own cause, its own fix and its own number, so each one is priced and reported on its own. If they were one engagement, a service that was working would cover for a service that was not.
Which one do you need?
A rough triage, in the order these usually present:
- Impressions or coverage are falling, or a template stopped being indexed. That is your own site. Start with technical SEO.
- You rank fine, and buyers arrive saying an assistant recommended someone else. That is AEO. Position and citation are decided by different systems and one does not buy the other.
- Your content is good, your site is clean, and you still sit under older competitors. That is authority. It is the slowest of the four.
- You already have the asset and the list of publications, and nobody has time to write two hundred individual emails. That is outreach. It is the child of the authority work, never a substitute for it.
If two of those describe you, the diagnostic decides the order. We run the one that unblocks the others first.
What the four share
One decision sits underneath all of them: which crawlers may fetch you. Access is granted per engine by a named token, never by a wildcard. The tokens for training, for AI search and for ordinary search are not the same. Blocking the wrong one removes you from an engine’s answers while leaving the rest intact, and it does so silently. We check that before anything else, on every engagement, because every measurement taken while an engine cannot reach you is worthless.
The second shared thing is delivery. Findings arrive as pull requests against your repository, reviewed by your engineers. An audit that ends in a document changes nothing. A search engagement that ends at a document produces no result at all, however good the document was.
What we turn down
Ask which of the four you need and you get one recommendation, not a bundle. Bought links are off the table, as is content we would not sign our own name to. A position we cannot attribute does not go in the report. Traffic figures from generated answers we do not quote at all. No platform exposes that number. Filling the gap with an estimate is how agencies take credit for movement they did not cause.
The expensive one, for us, is the last. When a service’s constraint is fixed and the next bottleneck sits somewhere else, we say so and stop. We do not invoice another quarter of the same work.
What you get
- A diagnostic that names which of the four services your problem actually is, with the evidence that led to the call
- A crawler access report covering search and AI tokens together, because one robots.txt line can decide both
- A baseline for every number we later ask to be judged on, recorded before any work starts
- A scoped proposal for the one service we recommend starting with, bounded so you can buy it alone
- A shared keyword and question map that both the on-site work and the answer-engine work are generated from
- One monthly readout when more than one service is running, so the same movement is never reported twice
How it runs
- 01
Diagnose before selling anything
We look at Search Console, your crawl, your referring domains and a sample of generated answers before recommending a service. The four failures look identical from the outside, and buying the wrong one wastes a quarter.
- 02
Confirm access first
Each engine grants access with its own crawler token, and a CDN or WAF rule can refuse one while allowing another. Any measurement taken while an engine cannot fetch you is invalid, so this is checked before a baseline is recorded.
- 03
Start with the binding constraint
Authority work on a site that cannot be indexed is wasted, and so is quotable copy an engine never fetches. We run the one thing that unblocks the others, and say plainly what we are leaving until later.
- 04
Report against that service's own number
Each service is judged on the measurement it controls, not on traffic. Traffic moves for reasons none of them caused, and a shared metric lets a failing service hide behind a working one.
- 05
Re-diagnose before adding a second service
The constraint moves once the first one is fixed. Re-running the diagnostic is what stops an engagement from quietly becoming a retainer that nobody can point at a result for.
Questions we get asked
- Can we just buy all four at once?
- You can, and in most cases you should not. One of the four is usually the constraint, and running the other three in parallel makes it impossible to attribute what moved. We would rather sell you one service you can judge us on than four you cannot.
- How do I know which one we need?
- Roughly this. Falling impressions or coverage errors point at your own site. Ranking well while buyers say an assistant named someone else points at AEO. Good content sitting under older domains points at authority. Having the asset and the target list but nobody to write the emails points at outreach. The diagnostic exists because these overlap more often than the summary suggests.
- Is the AI work just search work with a new label?
- Partly. The honest test is whether the seller can show you a measurement instead of a screenshot. Three differences survive that test. Access is granted per engine by a named crawler token. Most of those crawlers fetch JavaScript files without ever running them. The outcome is an inclusion decision that varies from run to run, and not a stable position. The rest of that pitch is recycled search work.
- Will you guarantee a position or a citation?
- No. Be precise about why, because the usual version of this answer is wrong. Ads do run inside and directly beneath Google's AI Overviews and inside AI Mode. Microsoft places ads inside Copilot answers. What is not for sale is the citation. There is no placement fee, no submission form and no schema that puts your name into the generated text. The same holds for a search position. What we commit to is the work, the measurement, and telling you when it did not move.
- What if the problem turns out not to be visibility at all?
- Then we say so and point you at the growth hub instead. A company can be found perfectly well and still lose the buyer on the page they land on. Selling that company a search engagement is a slow way to lose a client.
- Who actually does the work?
- The same team. We are engineers who also do the growth work. The fixes arrive as pull requests against your repository, and not as recommendations for someone else to implement. Nothing here is resold to a white-label supplier.
- Do these services share a report?
- Yes, one readout per month covering whichever are running, each against its own measurement. Separate reports per service is how the same movement gets counted twice. It is the easiest way for an agency to look busier than it is.