A tool reports. An agency acts. That is the honest one-line version, and the useful detail is in which acts actually matter.
1. Decide what to skip
Every audit produces forty items. Four of them move the number. Software cannot tell you which four, because the answer depends on your category, your competitors and what you already have, and it is judgment rather than data.
This is the least visible thing an agency does and usually the most valuable. It is also the thing you can buy on its own, as a one-off audit, without an ongoing commitment.
2. Claim things
The highest-value work in this field is completing the citation sources your category runs on. Every one is a form, a verification step, often a phone call, and always a human with your business details. No subscription does this.
It is dull work and it is the reason many businesses appear in answers at all.
3. Get you into other people's content
This is the one that separates being read from being recommended. The sources that decide who gets named are overwhelmingly not yours: round-ups, comparisons, industry lists, community threads. In the category we measure, Reddit was cited 225 times across 608 answers, ahead of every agency website in the field.
An agency can create a reason for someone to mention you, pitch it, and follow up. Software cannot participate in a conversation.
4. Write something worth quoting
Some tools bolt on generation. What comes out is generic, which is the specific failure mode here: the point is to say something a model finds worth lifting, and generic text is what a model already has infinitely much of.
Whether an agency does better depends entirely on the agency, and it is worth asking to see three pages they wrote for a client in a category you understand.
5. Notice when the strategy is wrong
A tool reports the number going nowhere. It does not tell you that the reason is a competitor's comparison page dominating the answer, or that your listings are inconsistent across two sources, or that the prompts you chose are not the ones your buyers ask.
Someone has to look at the answers themselves and read them. That is a person's job.
6. Be accountable
A tool cannot be wrong. When the number does not move, the dashboard is still correct and nobody owes you an explanation.
This is a real difference and it is most of what you are paying for. It is also why the contract term matters: accountability with a twelve-month lock and monthly reporting is weaker than it sounds.
What an agency cannot do either
Make an assistant recommend you.
There is no submission, no ranking factor, no relationship with OpenAI. What anyone can do is change what the models read and then measure whether the answers change. An agency describing it differently is describing something that does not exist, and that is worth catching in the first meeting.
They also cannot make your business into one that people want to recommend. If the reviews are bad, the AI answer will eventually reflect that, and it should.
Where a tool wins outright
Cost, speed to a first answer, and objectivity. A tool has no reason to tell you the number is moving when it is not.
The sensible arrangement for most businesses with a marketer is a tool plus a one-off audit: buy the judgment once, keep the execution in-house, and keep an independent measurement running. We recommend that to people it fits, which is more of them than our commercial interest would suggest.
Why number three is the one that decides it
Of the six, getting into other people's content is the one worth weighting heaviest, and the data says why.
We traced 608 measured AI answers back to the sources cited to produce them.
| Source | Answers citing it |
|---|---|
| reddit.com | 225 |
| aeoagency.org | 148 |
| youtube.com | 145 |
| linkedin.com | 134 |
| aeoengine.ai | 130 |
| thatware.co | 125 |
| techradar.com | 83 |
| en.wikipedia.org | 38 |
Every one of those beats every vendor website in the category, including the websites of the agencies that get named most often.
A tool can show you that you are absent from those sources. It cannot post in a community, pitch an editor, get you added to a round-up, correct a register, or build the relationship that gets you into next year's comparison article. That work is unautomatable by design, because the sources exist precisely to be resistant to automation.
This is also why "how many articles per month" is the wrong question to ask an agency. Articles are the part software could nearly do. The part it cannot do is the part you are actually paying for.
What that work produces
“Customers were asking ChatGPT and Gemini for product recommendations in our exact category, and we weren't even in the conversation. Within three weeks of launching Omni Eclipse's AEO strategy, we were the most-cited brand in the space - ahead of long-established brands with far more online presence.
PROCERT is a building certification firm that had grown entirely on referrals. Within a couple of weeks of the work going live, someone who had never met them inquired directly through the website. Fur Magic went from absent to the most-cited brand in their category in three weeks, ahead of long-established competitors.
Both published with the starting number, the end number and the period: PROCERT and Fur Magic.
A worked example of the gap
The same business, one month apart, with and without the execution.
With the tool alone. A dashboard showing thirty tracked questions, named on two, cited on nine, with a list of eleven sources they are missing from. Everything on that list is accurate and useful. Nothing on it has been done, because the marketer had a product launch that month.
What the eleven sources actually required. Two needed a phone call to correct an entry created by someone who left the company. One needed proof of registration uploaded. Three needed an account created and verified by post. One was a community where an account with no history would have been removed on sight. Four were straightforward form-filling.
What that adds up to. Roughly eleven hours of unglamorous, interruptible, low-status work spread across five weeks, most of it waiting on someone else.
Why it does not get done in-house. Not difficulty. It has no deadline, no visible output, and it competes with things that do.
That is the honest description of what an agency is for in this channel. Not insight. Follow-through on eleven boring things across five weeks.
See which questions name you
A free AI visibility audit on your own buying questions, across all five answer engines.
Book your audit
Ashur Homa
Built and scaled a digital brand to $100M+ in sales with zero ad spend. Has helped businesses generate millions through AI go-to-market strategy. Leads growth at Omni Eclipse.
Connect on LinkedIn