I asked an AI which knowledge base to buy, 204 times. Slite came up first once.

Same question, two channels, opposite answers. So this one doesn’t get solved with more SEO. And your website can’t solve it either, because Gemini never opens one:

Zero searches out of 69. It answers from memory. It is less like a search engine and more like a friend who has read a lot and never looks anything up. To get into that answer you have to be written about elsewhere. Writing more on slite.com won’t do it.

One thing does work, and it’s worth knowing. Name Slite and Gemini looks you up and gets it right. Asked cold, it doesn’t. It says you have a free plan, removed in June, with an invented 50-document limit, and it prices a “Standard” plan at $8. Your plans are Basic at $10 and Pro at $20 per user, billed yearly. Three of six facts wrong, all in the no-search arm.

I read all 24 of your English customer stories and counted what each customer says was broken before they found you:

• 17 of 24 — “we can’t find anything”: scattered, siloed, hours searching.
• 5 of 24 — “our documents are out of date”.
• 0 of 24 — out of date as the only problem.

“It’s impossible to find a document in Google Drive.” — Airmeet

“People kept asking the same questions again and again, even though they were documented on Notion.” — Wuffes

Nobody wakes up angry about stale documentation. People search on the day they lose forty minutes hunting for one file. Google agrees: your four positioning phrases return no data at all in its own ad tool, while “knowledge base software”, “ai wiki”, “second brain” and “single source of truth” each run 10K–100K a month. “Self-maintaining” is on all 21 pages I checked.

So the closer a buyer gets to buying, the further you fall. The bottom row is someone describing their live pain in their own words:

Don’t change the product. Self-maintaining is why people stay. It’s the wrong sign to hang above the door.

89% of your organic traffic lands on one blog post, /learn/knowledge-base-softwares. Your homepage gets 0.8%.

These are people shopping, not reading up. Anyone who just wants a definition types “what is a knowledge base”. That is a different search, and you are nowhere on page one for it.

So here is your best case. Someone lands on the #1 result for your biggest commercial term, reads thirty minutes of comparison, decides they like Slite, clicks, and gets “Talk to our sales team.” Nobody halfway through comparing tools wants a salesperson.

The fix is one line of HTML: Start free — 14 days, no card beside every Book a demo, and again at the end of the Slite section. No pop-up. The page is working. They are not failing to notice you, they are being sent to the wrong place.

• Almost all of it rests on one page. If that page slips down Google, most of your organic traffic goes with it inside a week. What fixes that is a second and third page that rank. Not a bigger budget.
• Under 1% of visitors search for Slite by name, so everyone arrives cold from a category search. Glean’s brand search sits a full band above yours: 100K–1M a month against your 10K–100K, in Google’s own bands, from a younger company. That gap is distribution, not product.
• 123 transactional keywords produce 49 visits a month. Nobody types “buy knowledge base software”, so there is no bottom-of-funnel demand to win. The money is commercial intent, where you already rank first.

And the traffic you do have is the good kind. 80% of your keywords are people reading up, each bringing under one visit a month. The 16% who are shopping bring 9.4 visits each:

All percentages, on purpose. Outside tools guess at visitor numbers, and Semrush’s guess is nowhere near Google’s own. You have the real ones.

• The homepage’s only button goes to a sales form too. Same choice as on the blog post, just on a much smaller page — the homepage is 0.8% of your organic traffic, that post is 89%.
• “Start for free” never says what free means — no 14 days, no credit card, nothing about day 15. It’s all in the pricing FAQ. This is also how the market learned you have a free plan, which is what the AI repeats back.
• Six fields are marked required on the demo form, and only five actually are — Job Title carries a star it doesn’t need. The button says “Contact sales”.
• Your demo page has eight customer quotes. Three praise how fast support replies — which is not why anyone buys a knowledge base. The one about the real problem, “the search never worked so nobody could find anything”, is fourth. Better quotes already sit on your own customer pages.

Then the trial. Two emails on day one: at 63 minutes, invite your team; at 102 minutes, here’s what Slite is, from Katerina. So you ask for colleagues’ names 39 minutes before anyone explains the product. And the forwarded invite opens “Hi Paresh Shrivastava,” so the colleague gets mail addressed to a stranger. Swap them.

Where I could be wrong

• I don’t have your real numbers, so treat every third-party estimate as a shape rather than a total.
• Only Gemini was tested. ChatGPT and Perplexity could look different.
• Six answers per question: enough to trust 0 of 6 and 6 of 6, not to tell 3 from 4.
• The trial is one person, one workspace, one day.
• I can’t prove that being written about elsewhere moves an AI answer. That’s why part three is a bet and parts one and two aren’t.

Three questions that would settle most of this in ten minutes, and only you can answer them: what share of organic traffic lands on /learn pages versus product pages · trial-to-demo conversion on each · and where your last twenty closed customers first heard of you.