Yes, partly, and the part you can do alone is larger than most agency pages admit. One person with a couple of unhurried afternoons, more if you carry several locations, can make their business facts agree everywhere a machine might check, claim and correct the listings that matter, write service pages a stranger could actually use, open the door to the crawler that feeds ChatGPT's search answers, and run a dated check across five assistants to see what comes back. None of that requires a marketing department. All of it counts.
Here is the part that rarely makes the brochure. The rest of the job does not have an end. Getting your business correctly identified is a task, and tasks finish. Staying in the answer is a standing job, closer to the monthly books than to painting a room.
Some of this work has a finish line. The rest only has a calendar. So the honest question is not whether you are capable, because you almost certainly are. It is which of those two things you are looking at, and whether the second one will still be happening next March.
What follows is the split as plainly as we can draw it: what genuinely finishes on your own, what that work buys and what it does not, where doing it alone stops paying, how to tell which side of the line you are standing on, and what to keep doing forever if you decide to stay the person who does it.
The Part You Can Finish Yourself
Start with identity, because it is the cheapest work with the largest effect and it has a genuine end point. Machines are trying to answer a boring question before they can answer an interesting one: which business is this, and is this the same business as that one. Every place your trading name, address, phone number, hours or service list disagrees with itself is a small argument a system has to settle without your help.
Google is unusually direct about this. Its help documentation says a Business Profile is built from four sources: publicly available information such as crawled web content, licensed data from third parties, users who contribute factual information including business owners who claim profiles, and information from Google's own interactions with the place. Your edits are one of the four. That is why we call it a record rather than a form, a distinction we go into further in what a Google Business Profile actually does for ChatGPT.
Next, open the door. OpenAI's crawler documentation says OAI-SearchBot is used to surface websites in search results in ChatGPT's search features, and that sites opted out of it will not be shown in ChatGPT search answers, though they can still appear as navigational links. To help a site appear, OpenAI recommends allowing OAI-SearchBot in robots.txt and allowing its published IP ranges. That is a quick check on a file you already own, and it is worth running before anything else, because a blanket block set years ago and never revisited is silent.
Then write the pages. A service page that names the situation, what is included, where your work stops, who is responsible and how pricing works does more for a stranger than any technical file, and you are the only person who can write it honestly. The full version of that argument is in what to put on your website so AI understands the business.
- Make the facts agree. Name, address, phone, hours and services, identical on your site and every listing. Fix the disagreements before adding anything new.
- Claim the records beyond Google. Bing and the major directories are separate filings from your Google profile, and keeping them right costs repetition rather than thought.
- Check robots.txt. Confirm you are not blocking the crawler that puts sites into ChatGPT search answers.
- Write one honest page per service. One page per thing you actually sell, in the words a buyer would use asking for it.
- Run a dated self check. Five buyer questions, no brand names, written down with the engine, date and exact wording. The method is in how to check whether ChatGPT recommends your business.
What That Work Actually Buys You
It buys eligibility. That is a real purchase, and it is not the same purchase as being recommended.
Google's guidance on optimizing for its generative features, updated on July 10, 2026, says that using products like Merchant Center and Google Business Profiles can help your products and services be visible in both AI responses and other Google Search results. Its documentation on AI features is blunter about the technical side: to be eligible as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to appear in Search with a snippet, and there are no additional technical requirements. Same page, same paragraph, Google states you do not need new machine readable files, AI text files or markup, and there is no special schema.org structured data you need to add.
Read that twice before you buy anything. A large amount of what is currently sold as AI optimization is a technical product solving a problem Google, on its own surfaces, says does not exist. If a vendor's entire offer is a file, a plugin or a markup package, Google has already answered them in public. Keep the money.
What the identity work cannot do is make anyone prefer you. It gets your business correctly recognized, which is the price of entry, not the win. As we put it in our own copy: AI can't feel your reputation in a room. It can only work with what it can verify. The tidy listing proves you exist and operate. It says nothing about why a buyer with a complicated situation should be sent to you rather than to a competitor with a thinner record and a louder website.
So the first honest checkpoint is this. If your self check showed you missing entirely, the identity work often moves it. If it showed you named, described roughly right, and still not the one being recommended, the gap usually sits in evidence rather than hygiene, and more listing work will not reach it.
Where Doing It Alone Stops Paying
Three parts of this job have no completion state, and they are the three that decide recommendations.
The first is corroboration you do not own. Your website is where you say things about yourself, and every competitor has one. What moves a system from confirming that you exist to treating you as a credible answer is agreement from places you do not control: independent coverage, business citations, reviews, credible mentions in the places your industry actually reads. Getting one is a good afternoon. Getting a steady supply is production work with relationships attached. That is the piece that quietly defeats most capable owners.
The second is cadence. One excellent page written in a quiet March does not survive a busy June. Google's own guidance asks for content that does not simply recycle what is already on the internet, and points at unique expert or experienced takes that go beyond common knowledge. That is exactly the material sitting in your head and nowhere else, and it is also the material that gets postponed every time a client emergency lands. The version of this that does not eat your week is described in monthly thought leadership without the homework.
The third is measurement done the same way for long enough to mean something. A single answer proves very little, because these systems vary between runs and between surfaces. Google says plainly that AI Mode and AI Overviews may use different models and techniques, so the responses and links they show will vary. A reading requires fixed questions, repeated runs, several engines and a date on every row, which is the discipline set out in how to measure whether AI visibility work is working.
There is a scale effect worth seeing clearly, and it is a map rather than a verdict. In one large vendor study, Ranqo analyzed more than 100,000 AI answers across 100-plus brands between March and May 2026 and found household names appearing in 73% of relevant unbranded answers, established mid-market brands in 44%, and small or niche brands in 11%. It is one company's study of its own market, so hold it loosely. The useful reading is not that small firms lose. It is that the questions the household names already own are the wrong questions to fight over. The specialist's move is to become the clearest available authority on the narrower ones nobody large has bothered to answer well.
How to Tell Which Side of the Line You Are On
This decision gets made badly because it gets framed as a judgment on the owner. It is not. It is a scheduling question with a commercial answer, and four plain checks will settle it faster than any sales call.
- Have you kept an unglamorous monthly task alive for a full year? The books, the invoicing, a newsletter that still goes out. If yes, cadence is genuinely available to you. If every such task has quietly died, this one will too, and knowing that is not a character flaw. It is data.
- Is your actual judgment written down anywhere a stranger can read it? If the distinctive thinking only exists in consultations, the bottleneck is capture, not publishing, and buying more pages will not touch it.
- What did the self check show? Missing usually points at identity. Named but generic, or named and passed over, points at evidence.
- Does anything outside your own domain confirm what you claim? If the honest answer is no, that is the work, and it is the least DIY-friendly part of the list.
One more thing worth doing before you decide, because it is free and it is yours. Microsoft opened AI Performance reporting in Bing Webmaster Tools on February 10, 2026, showing how often your content is cited in generative answers, which URLs are referenced, and how that activity changes over time across Microsoft Copilot, AI summaries in Bing and selected partner integrations. Microsoft is careful about what it means: the counts show how often pages are cited, not page importance, ranking or placement, and the data is a sample. Read with that caution in mind, it is a genuine signal you can watch inside a tool you register for yourself.
If three of those four checks say the standing work will not happen, hiring stops being an admission of failure and becomes arithmetic. If they say it will, keep going, and the next section is the routine to keep.
If You Decide to Keep Doing It Yourself
Plenty of good businesses will read all of the above and correctly decide to run it themselves. That is a legitimate answer, and it works when it is small enough to survive a bad month. Four things, held for two years, will outrun most of what gets sold as a project.
- Every quarter, run the same five questions the same way. Same wording, same engines, no memory of you, written down with the date. Comparable beats thorough.
- The same week a fact changes, change it everywhere. New hours, new address, new service, new person. Drift is what undoes the identity work, and it is silent.
- Once a month, publish one answer you already gave out loud. A question a client asked, answered properly, with your reasoning visible. Twelve of those in a year is a body of evidence.
- Once or twice a year, get one thing said about you somewhere you do not own. An industry publication, a professional association page, a real review from a real client. Slow beats absent.
Keep the boundary in view while you do it. Nobody outside the model companies controls what an assistant says, and any offer that promises you a fixed place in an answer is selling something it cannot deliver. What you control is whether the evidence exists, agrees with itself and can be checked in the couple of seconds a machine gives you. That part is entirely buildable, by you or by someone you pay, and the choice between those two is the only thing this article has been about.
If you would rather see the current picture before deciding anything, our free Recommendation Check asks the questions your buyers ask across five engines and scores your visibility across 109 checkpoints in eight zones. It takes about an hour to present, it costs nothing, and if the report shows you are already in good shape you keep it and we shake hands.
Want to Learn More?
Probably Genius builds and operates the recommendation layer around expert-led businesses, so what an assistant can verify actually matches the work being done. That is the job inside our done-for-you program: the expertise on the record, structured so machines can read it, corroborated well beyond a single listing, and measured across five engines every month. The published methodologies show how it runs, including the 109-Point Diagnostic behind the free check.
You're probably a genius at what you do. We make sure AI gets the memo.