60 of the AI tools we list have a homepage title that is only their own name
If you are launching an AI tool this year, you are launching into a category where nobody can name your competitors either. That is the actual problem: not that the market is crowded, but that almost everyone in it is invisible to anyone who does not already know their name.
Here is one measurable way that happens.
On 10 August 2026 we read the <title> and meta description of every homepage in this catalogue — 701 listings, 673 of which answered us.
What we found
| Listings | |
|---|---|
| Readable homepages | 673 |
— no <title> element at all | 8 |
| — title is the brand name and nothing else | 60 |
| — title over 60 characters, so it truncates | 125 |
| — no meta description | 41 |
| — description over 160 characters, so it truncates | 170 |
| — description under 50 characters | 37 |
The 60 that only say their own name
Ollama's homepage title is Ollama. CrewAI's is CrewAI. Browserbase's is Browserbase. So are TensorFlow's, Cerebras's, GitButler's, Artbreeder's and AgentGPT's. One site manages Haystack | Haystack.
These are not amateur projects. Several are the best-known thing in their niche. And the title tag is still the single strongest on-page signal about what a page is for — it is the blue line in the result, and it is what a search engine matches against the words somebody typed.
A title of Ollama can match one query: ollama. Which is only typed by people who have already heard of Ollama. Every person searching run llama locally or local llm mac is looking for that product and being shown someone else.
The fix costs nothing and takes a minute: Ollama — run large language models locally. The brand stays, and now the page also says what it is.
The 125 that get cut off
Search results show roughly 60 characters of a title. A quarter of the readable catalogue writes past that, and what falls off the end is whatever came last.
Two real examples from this dataset:
AI Image Upscaler - Enlarge & Enhance Your Photos for Free - Upscale.media(74 characters — the brand is what gets cut)LLMStack | AI Agents in Minutes | No-code AI App Builder | LLMStack(67, and the brand appears twice)
This is a less serious problem than the bare brand name, because a truncated title still matches the query — the words are all there for the engine even when the reader cannot see them. But it means the part you chose to end on is the part nobody reads.
Descriptions: 41 missing, 170 too long, 37 too short
The meta description does not affect ranking. It affects whether the person clicks, which is the part that actually matters to you.
41 have none, so the engine writes one by lifting a sentence from the page. Sometimes that is fine. Sometimes it is your cookie banner.
170 run past 160 characters and get cut mid-sentence. One in this set is 346 characters long — more than twice what will ever be shown.
37 are under 50 characters. These are not failures, and some are better than the long ones: Read research papers 10x faster is thirty-one characters and does more work than most of the 200-character entries. Short is only a problem when it is short and vague.
The eight with no title element
Eight homepages have no <title> in the HTML a crawler receives. A search engine has to invent one, usually from a heading or the domain.
We nearly reported fourteen. Our first pass read the first 200KB of each page on the reasonable theory that <head> comes first, and Cohere, Databricks and Granola came back as having no title — their pages are 673KB, 838KB and 351KB, with enough inlined script above the tag to push it past our cut. That was our bug, not theirs. Reading the whole document moved six sites out of this bucket and fourteen out of the missing-description one.
Why this is a founder's problem specifically
Nothing here requires a rewrite, a redesign, or an agency. Every item is one line of HTML, and every one of them is a line most teams wrote once, at the beginning, before they knew what the product was for.
If you want to know how your own site reads to a machine, paste your URL into our submission form. It runs seven checks — reachability, title, meta description, og:image, logo, whether we already list you, and whether you link back — and returns a specific fix for each one. It does not require an account to run, and you do not have to submit anything afterwards.
What we are not claiming
One request per homepage, from one machine, on one day, reading the HTML as delivered. 28 sites refused us and are excluded rather than counted as failures.
We read the static document deliberately, because that is what a crawler indexes before it runs anything — but a title injected by JavaScript would show up here as missing, and we cannot tell those two cases apart from the outside.
"Brand name and nothing else" means the title, once the product's name and any separator characters are removed, has three characters or fewer left. That rule counts Haystack | Haystack as bare, which is right, and would count a title like Notion – AI as bare, which is arguable. The 60 were eyeballed after the fact and we did not find one we disagreed with, but the definition is the definition.