Foundations
I listed on 20 directories. What next?
By Arnav Mukherjee, founder of TofuBofu · July 31, 2026
TL;DR
A listing says you exist. Engines are trying to work out who is good, and an empty profile answers a question nobody asked. The directory push was the cheap half and you have finished it. The next four moves, in order: get reviews on the two or three platforms your category is genuinely evaluated on, go where your buyers actually discuss the problem, write the comparison content the directories cannot write for you, and verify what engines are citing rather than assuming there is a canonical list. Reddit is roughly 12% of ChatGPT citations. Directory number 21 is not.
This question turns up constantly in founder communities, usually phrased with visible frustration. Someone spent a weekend, or paid a submission service, and got themselves onto twenty or fifty directories. The listings are live. The badges are on the site. And when they ask ChatGPT who the best provider in their category is, they are still not in the answer.
The frustration is fair, because the advice they followed was not wrong exactly, it was incomplete in a way that made it feel like a finished job. Directory submission is genuinely worth doing. It is just the floor, and almost every guide stops at the point where the real work begins.
Why a listing does almost nothing on its own
Think about what an engine is doing when a buyer asks for the best provider of something. It is assembling a shortlist, and to justify putting a name on that shortlist it needs material it can point at. Now consider what a bare directory listing actually contains: a company name, a URL, a category tag, maybe a two-line description you wrote yourself.
There is nothing in that record that helps answer "who is best." It confirms you exist and that you claim to be in a category. Both were already true. The engine cannot quote it, compare it, or use it to rank you against anyone, so it mostly does not.
This is the distinction that the twenty-directory weekend misses. There are two kinds of third-party presence and they behave completely differently:
| A listing | A citable signal | |
|---|---|---|
| What it proves | You exist | You are good at something specific |
| Who wrote it | You did | Someone who is not you |
| Can an engine quote it? | Not usefully | Yes, and it does |
| Effort to obtain | Minutes | Weeks, and a real customer |
| Examples | Bare profile, badge, index entry | Reviews, a comparison writeup, a community thread |
Twenty listings gets you twenty entries in the left-hand column. The reason it feels like it should have worked is that in classic SEO it partly did: links were the currency, and a link is a link whether or not anyone reads the page. AI citation runs on a different currency. It wants substance it can attribute, and it prefers sources that other sources also trust.
There is no canonical directory list, and that matters
The premise behind a twenty-directory push is that there is a known set of sites AI engines consult, and getting onto all of them is the job. Our own data pushes back on that, though with a caveat I want to state before the finding rather than after it.
Across the scans where engines returned identifiable sources, we recorded 129 citations spanning 109 distinct domains. That is close to one citation per domain: almost nothing repeated. The most frequently cited single source was Clutch, at five appearances, under 4% of the total. The rest was a long, thin tail of category-specific sites, individual company pages and one-off references.
The caveat is that 129 is a small number and I would not build a strategy on it alone. It is enough to support a modest claim and no more: cited sources are far more dispersed and far more category-specific than a universal directory checklist implies. The larger published research points the same way on the sources that do repeat. Profound's analysis found Reddit accounts for roughly 12% of ChatGPT citations, well ahead of Clutch at around 2.4%, and Kevin Indig's work with G2 found G2 holding 22.4% share of voice specifically for software queries. Note what those three have in common: they are all places where people write about companies, not places that list them.
Find out which sources AI cites in your category
A free scan runs your real buying questions across six engines and shows the full verbatim answers, including which sources each engine leaned on and which competitors it named instead of you.
Run your free scanThe four moves that come next
In order. The first one is worth more than the other three combined, and it is the one people skip because it requires asking customers for something.
1. Get reviews on the two or three platforms that count in your category
You are already listed. Now make the listings evaluative. Pick the platforms where your buyers actually compare vendors, usually G2 and Capterra for software and Clutch for services, and work your customer base for genuine reviews. The relationship is measurable: roughly 2% more AI citations associates with 10% more reviews. Ask at the moment a customer tells you something worked, not in a quarterly batch email, and never buy or fabricate them, which violates every platform's terms and gets wiped.
2. Go where the discussion is, and be useful rather than promotional
Reddit is around 12% of ChatGPT citations, more than any review platform, because it is where people ask each other what actually works. The move is not to post about your company. It is to answer questions in your area of expertise properly, from an account with a history, so that your name appears in a context an engine reads as credible. This is slow and cannot be outsourced convincingly, which is exactly why it is worth something.
3. Write the comparison content directories cannot write for you
Directories hold structured facts. They do not explain why a buyer with your specific constraints should choose you over a named alternative, and that explanation is precisely what an engine needs when it assembles a shortlist. Write honest comparison and use-case pages that name real alternatives and real trade-offs. This is content you fully control, it takes days rather than quarters, and it is the fastest route into narrow questions where incumbents are too generic to be the right answer.
4. Verify what your engines actually cite, then stop guessing
Run your real buying question through each engine several times and record which domains come back as sources. Several times matters, because repeated identical queries return substantially different answers and a single run will send you after the wrong sites. What emerges is your category's actual source list, which is usually shorter and stranger than any published directory roundup, and it tells you where the remaining effort belongs.
What not to do next
Do not add directories 21 through 100. The return per additional listing was already low at twenty and falls from there. Bulk submission services in particular put you on sites engines do not consult, generate a burst of near-identical entries that pattern-matches to manipulation, and eat time that belongs to the four moves above.
Do not chase removals from useless listings. A thin entry on an irrelevant site is mostly inert. The exception worth acting on is any listing carrying wrong information about you, because incorrect facts on third-party sites propagate into AI answers and are considerably harder to correct once an engine has started repeating them.
Do not conclude directories were a waste. They were the floor and you now have it. Being absent from the places your category is evaluated is a genuine problem, and you have fixed it. The mistake was only in expecting the floor to be the whole building.
The uncomfortable through-line is that everything which actually moves AI visibility requires another person to say something about you. Reviews need customers. Community credibility needs a history. Comparison content needs you to be honest about alternatives. None of it can be completed in a weekend, which is exactly why the twenty-directory weekend is so appealing and why it did not work.
Frequently asked questions
I listed on 20 directories and AI still does not mention me. Why?
Because a listing is a record that you exist, not evidence that you are good, and engines weight evidence. An empty profile on a directory says your company has a name, a URL and a category. It contains nothing an engine can quote when a buyer asks who is best. The listings that move AI visibility are the ones carrying reviews, ratings, detailed service descriptions and comparison context, which is a small subset of any twenty-directory push. The submission was the cheap half. The half that matters is filling the two or three profiles your category is actually evaluated on.
Do directory listings help AI visibility at all?
They help as a floor, not as a lever. Being absent from the directories your category uses is a real problem, because it removes you from a source engines consult and from the comparison sets they assemble. Being present with a thin profile fixes the absence and adds very little beyond it. Research from Profound found brands present on four or more platforms are 2.8x more likely to be cited, so breadth is genuinely valuable, but the platforms in that count are ones carrying substantive content about you rather than bare entries.
Which directories actually matter for AI citations?
The two or three your buyers actually use, which differ sharply by category. For B2B software that usually means G2 and Capterra, where G2 alone has been measured holding 22.4 percent share of voice for software queries. For services firms Clutch carries more weight, at roughly 2.4 percent of ChatGPT citations in Profound's analysis. Community platforms outrank both: Reddit accounts for around 12 percent of ChatGPT citations, more than any review site. The practical test is to ask which sites appear when you run your own category question through an engine, then invest in those rather than in a list of twenty.
Is submitting to 1000-directory bulk lists worth it?
No, and it can be actively counterproductive. Bulk submission services put you on low-quality sites that engines do not consult, produce a burst of near-identical listings that reads as manipulation, and consume time that would be better spent on the few profiles that carry weight. The value of a directory to an AI engine comes from that directory being a trusted source, and the vast majority of the thousand are not. Depth on three platforms beats presence on three hundred.
What is more important than directories for getting cited by AI?
Reviews on the platforms you are already listed on, and genuine discussion of your company in places buyers talk. Reviews are the highest-return next move because they convert a static listing into evaluative content, and analysis by Kevin Indig with G2 found roughly 2 percent more AI citations associated with 10 percent more reviews. Community presence matters more than most founders expect, since Reddit alone is around 12 percent of ChatGPT citations. Both of those beat adding directory number 21.
How do I know which sources AI is actually citing in my category?
Check directly rather than assuming, because it varies more than people expect. Open each engine, ask the exact question a buyer would ask before they knew you existed, and record which domains appear as sources. Do this several times per engine, since repeated identical queries return substantially different results and one run will mislead you. The pattern that emerges is your actual target list. In our own scan data the cited sources were strikingly dispersed, with very few domains repeating across scans, which is a warning against assuming a universal canonical directory set exists.
Should I remove my listings from low-quality directories?
Generally it is not worth the effort unless the sites are genuinely spammy. A thin listing on an irrelevant directory is mostly inert: it does not help, but it rarely hurts enough to justify the time to chase down removals. The exception is directories that publish inaccurate information about you, because incorrect facts on a third-party site can propagate into AI answers and are much harder to correct once quoted. Prioritise correcting wrong information over deleting merely useless entries.
Sources and further reading
- TofuBofu first-party scan data: 129 identifiable source citations across 109 distinct domains, from scans run June to July 2026. Small sample, reported as directional evidence of dispersion only
- Profound research: Reddit around 12% of ChatGPT citations, Clutch around 2.4%, and brands on four or more platforms 2.8x more likely to be cited
- Kevin Indig and G2 on reviews and AI citations: G2 holds 22.4% share of voice for software queries, and 10% more reviews associates with roughly 2% more AI citations
- SE Ranking AI search study: 71% of ChatGPT-cited pages use structured data