sim.ai

Read 3 days ago, the first 100 pages. Read it again

Top 25%of the 1,062 sites in San Francisco, California, at #152

It scores 87. Gemini cites it 4th of 4 sources when asked “What are the best AI assistants?”. See the answer

A score says pages are well formed. It does not say an assistant cites the site.

Fix this first

Add a tap-to-call link

87 now93 with itTop 10% from 90

#46,938to#10,77736,161 places up100 of its 100 pages fail it. 1 in 3 ranked sites pass it.

Shop Ready, 1 product graded, 4 of 12 fields

Where it ranks

3 points take it into the top 10%.

Open the leaderboard

Beside it on the board

  1. #46,936oregondiscountwindows.com87
  2. #46,937nycgaragedoorsandgatesrepair.com87
  3. #46,938sim.ai87
  4. #46,939tuning-art.com87
  5. #46,940stevekoophotography.com87

7,567 sites score 87. A tie goes to the site with faster, leaner pages that say more and keep improving.

To the top 10%

  1. 87Now#46,938
  2. 93Add a tap-to-call linkTop 10%

1 fix. Top 10% starts at 90.

Where it stands

Each stage against the middle site. It trails on Trust.

Access35 points93
Middle site
91
Ahead by
2

Worth the mostLink straight to the final address+1

Understand30 points88
Middle site
59
Ahead by
29

Worth the mostAnswer the first question right under it+2

Trust20 points96
Middle site
98
Behind by
2

Worth the mostEach fix here is worth under a point

Act15 points57
Middle site
57
Level with it

Worth the mostAdd a tap-to-call link+6

Against a typical site

The readThis siteTypicalAgainst typical
Pages read1004456 more than typical
Links a page64134.9× typical
Words a page1,9483944.9× typical
Rules failing somewhere17143 more than typical
Half its pages answer in108 ms210 ms49% faster than typical

This siteWorse than typicalTypical, the middle ranked site, worked out each hour

The work, worth the most first

17 fixes. The first 3 are worth the most; each group says the places it lifts the site, counted from where every ranked site scores now.

Open the action plan
FixIts pages that fail it, fewer is betterSites passing
Worth 6 points36,161 places up
Add a tap-to-call linkAct100 of 1001 in 3
Worth 2 points14,714 places up
Answer the first question right under itUnderstand43 of 1001 in 121
Worth 1 point7,847 places up
Link straight to the final addressAccess99 of 1001 in 2
Show 14 more fixes
Worth under a point each
Put figures in tablesUnderstand24 of 1001 in 28
Phrase headings as questionsUnderstand13 of 1001 in 10
Add depth to thin pagesUnderstand13 of 1003 in 7
Put the text in the HTMLAccess8 of 1002 in 3
Show sources on guidesTrust7 of 1001 in 2
Add lists or tablesUnderstand10 of 1001 in 2
Show when guides were updatedTrust4 of 1003 in 5
Match structured data to the pageUnderstand4 of 1003 in 7
Fix description lengthsUnderstand7 of 1001 in 11
Describe imagesUnderstand7 of 1002 in 7
Name the publisherTrust4 of 1003 in 4
Bring deep pages closer to homeAccess5 of 1006 in 7
Fix title lengthsUnderstand5 of 1001 in 6
Set a share imageUnderstand2 of 1002 in 7

How AI reads this site

It reads mostly as software and SaaS. 3 changes would make it plainer.

It reads as

Software and SaaSIt says it is an organization in San Francisco.

  1. None
  2. Loosely
  3. Split
  4. Mostly
  5. Clearly
  1. Closest lines of workAgainst the closest
  2. Software and SaaS
  3. Agencies and studios
  4. Real estate

What an assistant reads first

Home pageWhat it says
TitleThe AI Workspace | Build, Distribute & Govern AI Agents | Sim
DescriptionSim is the open-source AI workspace where companies build, distribute, and govern AI agents. 270+ integrations, every major LLM, permission groups, spend limits, and self-hosting.
First headingBuild and governevery AI agent in one place
OpeningSim is the open-source AI workspace where teams build, deploy, and manage AI agents for their organization. Connect 270+ integrations and every major LLM, then govern access, spend, data, and deployment from one place. Build visually, conversationally, or with code, and run Sim…

3 changes so it reads plainly as what it is

  1. Say what you do, for whom and where

    Its pages read only partly as Software and SaaS. An assistant matches a site to a question by what its pages plainly say.

    Agent instructions

    On sim.ai, rewrite the home page title and its first sentence to say plainly what the business does, for whom and where, so the site reads as Software and SaaS. Keep the title under 60 characters and lead with the service, not the name.

  2. Rewrite the description

    Its description is 179 characters and is cut off in results.

    Agent instructions

    On sim.ai, write the home page a meta description of one or two sentences, 50 to 160 characters: what the business does, for whom, and where.

  3. Name San Francisco in the first words

    The site is in San Francisco, and its title, description, first heading and opening never say so. Questions to assistants often name a place.

    Agent instructions

    On sim.ai, name San Francisco in the home page title or first sentence, and in the description, where it reads naturally.

  4. Read it again

    mttr reads the site again and says how it reads now.

An estimate made by mttr against 16 lines of work, from its home page before any script runs. It is not what any assistant says about the site. Rethink the direction with us.

Gemini cites sim.ai 4th of 4

mttr picked the question from sim.ai’s home page and asked it as a person would.

  1. GeminiOctober 22, 2026Cited 4 of 4

    What are the best AI assistants?

    Gemini

    The "best" AI assistant depends entirely on what you want to use it for. The market has evolved past simple chatbots into specialized tools, workspace integrations, and autonomous agents. The top AI assistants are categorized below by their primary strengths: ---.

    1. 1
      ChatGPT (OpenAI)
      • Best for Everyday tasks, brainstorming, casual conversation, and multimodal features (voice, image…
      • Why it’s great It remains the most well-rounded and widely adopted conversational AI.
    2. 2
      Claude (Anthropic)
      • Best for Long-form writing, software coding, and analyzing massive documents.
      • Why

Asked with only the question sent; answers vary by day, place and person. Passing a rule and being cited go together here, a correlation, not proof. A score says pages are well formed. It does not say an assistant cites the site.

mttr has read 2 of the 3 sites cited.

Fix sim.ai with your agent

Your agent makes the fixes. mttr checks each one. Free.

In the site’s repositorynpx skills add subinternal/mttr
Then tell your agentCheck sim.ai with mttr and fix what’s worth the most.

Claude Code, Codex, Cursor or any agent that reads AGENTS.md.

Claude Code, onceclaude mcp add --transport http mttr https://mttr.si/mcp
It needs a free key
Another agent
Read it againChanged something? A site is read at most once every ten minutes.