directree does not show composite scores - no 8.7/10, no star ratings, no "Ease of use: 9.2." That number looks authoritative, but the precision is almost always fake: a proprietary weighting of noisy social signals dressed up as measurement. We think that misleads buyers and actively harms the founders it claims to rank.
This post explains exactly why -- and what we show instead.
What "fake precision" actually means
Precision, in the technical sense, means the number of significant figures you claim. When a scale reads 82.4 kg, the decimal implies your scale can distinguish between 82.3 and 82.5. If it can't -- if it just rounds to the nearest kilo and the decimal is meaningless noise -- that decimal is fake precision. It makes the reading look more reliable than it is.
Software scores are almost always fake precision.
Think about how a 8.7/10 rating gets calculated on a large review platform. A vendor collects reviews -- often prompted by an email blast to customers who just had a good support interaction. Some review sites allow incentivized reviews (gift cards, free months) as long as they disclose it. The reviews get weighted by recency, reviewer verification level, company size, and a handful of other factors that vary by platform. Then this composite gets displayed as a single number to two decimal places, next to products with very different review counts, very different incentive structures, and very different user bases.
The 8.7 is not a scientific measurement. It is an aggregate of noisy social signals, filtered through a proprietary weighting algorithm that you can't audit. Displaying it to one decimal place implies a precision that doesn't exist.
That's a problem for buyers. And it's a different kind of problem for founders.
Why founders get hurt by fake scores
If your product has 12 reviews and a competitor has 4,000, you're not competing on the same scale. But both of you might display a star rating or a composite score. Yours might actually be higher -- fewer reviews, more recent, less incentivized noise -- but it'll look thinner, because the count matters to how people interpret the number.
A new product listing on most review-aggregator directories faces a choice: either show up with "Not enough reviews yet" (which reads as unproven), or gather reviews fast -- which means incentivizing them, which means noise. Neither is honest. Both push founders toward gaming a system that's supposed to help buyers.
There's another problem, one we care about a lot at directree: AI.
When ChatGPT, Perplexity, or Google's AI reads a page and finds "Acme CRM: 8.7/10," it tends to repeat that number. It treats it as a fact. In AI-generated responses, that score gets cited alongside genuine facts like "founded in 2019" or "integrates with Slack." The AI doesn't know it's a composite artifact of a review collection methodology. It just sees a number and reports it.
That means a fake-precision score doesn't just mislead the buyer who reads it on the directory page. It leaks into the AI knowledge graph and gets recycled as apparent fact for months, sometimes years. We think that's genuinely harmful. It's one of the reasons we took such a strong position here when designing directree.
What we built instead
directree doesn't do scores. No composite ratings. No "8.7 out of 10." No star ratings derived from a review count we don't have.
What we do instead is show three clearly-labelled tiers of information, described in detail in the honesty model post:
- Observed (green, monospace): things we can verify directly from the tool's website. Pricing tier, free plan, open source status, platforms supported, integration logos. These are stated as facts because they are facts.
- AI-inferred (rose, always labelled): summaries, strengths and weaknesses, best-for and not-for assessments. Useful -- we think they help buyers navigate -- but explicitly marked as AI interpretation, not editorial judgment.
- Founder-edited (gold): fields that have been claimed and corrected by a verified owner. The highest trust tier, because there's a real person accountable for the information.
None of this produces a numeric score. Instead, a buyer can see what a tool actually costs, whether it has a free plan, what the AI thinks its strengths are (and can discount that accordingly), and what the founder themselves says about who it's best for.
That's harder to rank-order. You can't sort a directory by "8.7 > 8.4." We think that's fine. Ranking tools by a fake-precision composite score doesn't actually help buyers pick the right tool for their situation. It helps marketers know which tools to push, and it helps platforms charge more for "Top Rated" badges.
The build-in-public part: this was a deliberate design decision made early
We made the call to skip numeric scores during the first week of designing directree, before any code was written. The question came up naturally: what would our "rating" look like?
The answer we kept coming back to was: a rating derived from what data? We don't have thousands of verified reviews. We have one source of truth per listing -- the tool's own website, plus AI enrichment. You can't produce a credible numeric score from that without faking it.
So we didn't. We went with provenance labels instead. And the more we thought about it, the more it became a positioning principle rather than just a technical limitation. The numeric score isn't something we're planning to add once we have more reviews. It's something we're actively choosing not to do, because we think it makes directory data worse.
One honest note: directree does show community upvote counts from our own users. That's a real signal -- it tells you which tools people in our community have found worth flagging. But it isn't a rating. A high upvote count doesn't mean "9.1/10 product." It means "a bunch of people who use directree thought this was worth an upvote." Different claim. Lighter claim. Truer claim.
Why this matters for the broader directory problem
Most software directories are either (a) review aggregators that need review volume to be useful, or (b) pure crawl-and-list tools that show you a logo and a name and nothing else.
The review aggregators have the fake precision problem. The crawl-and-list tools have the information vacuum problem. directree is trying to thread a third path: rich, structured, provenance-labelled data that's honest about what it knows and what it's guessing.
If you want to understand the broader category -- what directories are, how they work, what they're good for -- this piece on software directories in 2026 covers it. And if you've wondered whether sites like this are still worth listing on for SEO, there's some honest analysis there too.
The short version: we think honest structured data is better for buyers, better for founders, and better for AI systems that are increasingly using directory data to form opinions about software. The fake score is a shortcut that serves none of those audiences well.
FAQ
Why does directree not show star ratings? Star ratings typically aggregate user reviews. directree doesn't collect user reviews in the traditional sense -- our listings are built from observed web data, AI enrichment, and founder corrections. A star rating derived from that data would be fabricated. We show what we actually know, labelled by how we know it.
Will you ever add user reviews or ratings? Possibly, in a limited form. If we do, we'll label them clearly and won't aggregate them into a composite score. The label problem doesn't go away just because users are writing the reviews -- you still need to know whether a review was incentivized, how old it is, and how many there are before a rating means anything.
Doesn't this make it harder to compare tools? Harder to rank-order by a single number, yes. But we'd argue that "which tool has the higher fake score" isn't a useful question. Our compare pages and community surveys are better comparison surfaces -- structured, head-to-head, built from the same labelled data model.
What is the community upvote on directree? When logged-in users see a tool they find valuable, they can upvote it. The upvote count is a real signal of directree community interest, displayed openly. It isn't a quality rating. A tool with 1 upvote and a verified free plan is not "worse" than a tool with 50 upvotes that is AI-inferred only. The signals mean different things.
How does directree handle tools that haven't been claimed yet? The listing still shows all Observed data (from the crawl) and AI-inferred fields (clearly labelled). Unclaimed listings are honest about being unclaimed. The founder-edited tier simply doesn't appear until someone verifies ownership. This is a feature, not a bug -- you can see exactly how much of the listing has been verified by a human with skin in the game.