AI referral opportunity triage report _

When someone asks ChatGPT or Claude a question about Gridwell, the assistant frequently sends them to a help center article. The help center is large, 429 pages received AI traffic and 4,150 visits in total, and most of it is fine. This action is about eleven specific articles. Across the 42 days of clean data, 604 visitors landed on those eleven and 23 requested a demo, a rate of 3.8 percent. Five other help articles, using exactly the same page design, converted between 18 and 24 percent of their visitors. The page design is therefore not the obstacle. The difference is subject matter. Articles that walk someone through building a specific thing, such as a gallery view or a form, leave the reader wanting to try it. Articles that explain a concept, describe an admin setting, or cover an AI feature leave them with nothing obvious to do next.

The same page design converting well, for comparison:

1. Add a button at the end of the article that opens the product and starts the reader doing whatever the article just explained. On the view builder article that means Create your first interface, not a generic Book a demo in the site header. The reader has just learned how something works, so the next click should let them try it.
2. Fix the four largest pages first: contacting-gridwell-support (220 visits, zero demo requests), using-the-gridwell-mcp-server (96), getting-started-with-view-builder (90) and getting-started-with-automations (82). Those four are 488 of the 864 visits.
3. Handle contacting-gridwell-support differently. At 79 percent leaving immediately and zero demo requests, assistants are surfacing it as a general Gridwell contact page. Add a link to the product tour or plans above the support form, because many of these visitors are not existing customers and have nothing to contact support about.
4. Before building anything, look at docs/gridwell-plans. It converts at 23.8 percent using the identical page design, so whatever it does differently is the thing to copy.
5. Do not fix the three AI-feature articles (using-assist-in-gridwell, using-gridwell-ai-in-fields, using-the-gridwell-mcp-server) with a button alone. Action 4 shows they are also competing with other Gridwell pages for the same searches.

Account creation events from the warehouse. The gap between 3.8 percent and 20.4 percent holds across 16 pages and 1,383 visits, so it is not a small sample effect. The share of people leaving immediately supports it too: 69 percent on the weak group against 51 percent on the strong one.

Section 06 shows that Google's AI answer for "gridwell automations" quotes five help center articles and does not quote /platform/automations, which is the top ranked result. Help center articles are the pages assistants prefer, so this conversion gap sits exactly where the citations are landing.

Why 604 and not 2,902. Section 04 shows the whole Support pages category at 4,150 visits across 429 pages over 57 days. This action covers only the eleven articles that both carry meaningful traffic and convert under 7 percent, measured over 42 days. It is a subset of that category, not a different measurement of it. Expanding the Support pages row in section 04 shows the ten largest pages in the category, several of which convert well and are not part of this action.

Note on the window. Tracking for support.gridwell.example stopped recording on 11 August 2026 and has recorded nothing since. These figures cover 30 June to 10 August, 42 days rather than 57. The real weekly volume is higher than a 57 day average would suggest, so this is the most understated action on the list.

AI assistants send people to the privacy policy and the terms of service when answering questions about how Gridwell handles data, security and contracts. Over the eight weeks measured, 433 people arrived on those two pages and three requested a demo. They stayed 1.2 to 1.5 minutes on average, long enough to read the section they came for and leave. This is not idle traffic. Someone checking data handling terms is usually assessing Gridwell on behalf of a company. Neither page currently gives them anywhere to go afterwards.

1. Add one block below the legal text, not inside it: a short security and compliance summary linking to the trust materials, plus a single link to start a free demo request.
2. Do not put a Book a demo button in the middle of the document. These readers are checking terms. The /about page converts at 22.6 percent using a quiet link at the end rather than an interruption.
3. Include a link to plans or contact sales as well, because some of this traffic is procurement rather than an individual requesting a demo.
4. Judge the result against /about at 22.6 percent, the closest comparable page, rather than against the site average.

Account creation events from the warehouse, 30 June to 25 August 2026. Both pages sit in the Other Pages category, which converts at 6.6 percent overall, so these two are far below even their own group.

Spending 1.2 to 1.5 minutes on a page and then leaving, which 62 to 63 percent of these visitors do, is the pattern of someone who found their answer. There is currently nothing on the page for them to do next.

Of every page type on the site, articles convert AI visitors worst, at 8.0 percent, and 71.6 percent of those visitors leave without clicking anything. The problem is not spread evenly. Two articles, both about AI tools, lose more than 80 percent of visitors and hold them under a minute. A third article about project management software, built the same way, holds visitors nearly seven minutes and converts 16.7 percent. When two pages behave that differently from a third built identically, the cause is usually something specific to those pages, such as where the button sits or whether the opening paragraph answers the question completely. That is worth checking before anyone rewrites the content.

The third row is the comparison case. Same page design, same category, converts three times better with 30 points fewer people leaving.

1. Open /articles/ai-workflow-automation-tools and /articles/no-code-ai-tools and check where the button sits. At 0.5 and 0.9 minutes average time on page, a button below the fold will never be seen.
2. Check whether either page opens with a comparison table that answers the question completely in the first screen. Both rank for list style searches where a reader can leave satisfied without scrolling.
3. Compare against /articles/free-project-management-software, which holds people 6.8 minutes and converts at 16.7 percent, and identify the structural difference.
4. Only rewrite the content if button placement and page structure turn out to be fine. The evidence so far does not say the writing is the problem.
5. Use Mouseflow scroll and click recordings once available. They will show directly whether visitors ever reach the button, rather than inferring it from time on page.

Account creation events from the warehouse, 30 June to 25 August 2026. Articles carry the strongest search traffic value on the site and the only meaningful Beaconpath citation rates, so this traffic is earned and worth converting.

no-code-ai-tools has a 0.25 percent Beaconpath citation rate, the second highest of any Gridwell page in the tracked category, and the second highest AI search demand in section 03 at 220. It is winning visibility and losing the visitor.

These three categories look healthy at the category level, converting between 16.8 and 18.4 percent, which is why they did not appear in earlier versions of this list. Looking inside them changes that. Each contains a tail of pages that convert near zero while the category average is held up by a few strong performers. Across the three, 18 pages take 273 visits and produce 22 demo requests, a rate of 8.1 percent. Five of those pages converted nobody at all. The category average hid them.

Guides, 8 pages, 135 visits, 8.1 percent combined. Category average is 16.8 percent.

Solutions, 5 pages, 74 visits, 5.4 percent combined. Category average is 17.5 percent.

Templates, 5 pages, 64 visits, 10.9 percent combined. Category average is 18.4 percent.

1. Treat this as the same fix as action 1, applied to three more page designs. Each page needs a button that starts the reader doing the thing the page describes.
2. Start with Guides, which has the most traffic at 135 visits and the clearest pattern. /guides/build/connect-data-with-linked-records alone is 48 visits at 14.6 percent, and the two guides that convert nobody, view-builder-record-review and save-time-with-automations, both lose over 75 percent of visitors immediately.
3. For Solutions, look at whether these pages have any product path at all. /solutions/finance and /solutions/enterprise convert zero on 30 combined visits, and /solutions/hr loses 82 percent of visitors immediately. These read as pages built for a sales conversation rather than a self-serve demo request.
4. For Templates, the weak pages are individual template detail pages rather than the hub. /templates itself converts at 16.9 percent on 337 visits and needs no work. Check whether the weak template pages are missing the Use this template action that the strong ones have.
5. Copy from within each category rather than across. /guides/start/what-is-gridwell converts at 30.2 percent and /templates/event-planning at 31.0 percent, so a working pattern already exists in each section.

Account creation events from the warehouse, 30 June to 25 August 2026. Pages are included when they received at least 8 visits and converted under 15 percent, which is why the list is 18 pages rather than every page in those categories.

This action exists because a reviewer asked why Templates, Guides and Solutions were absent from the list. They were absent because the category averages looked acceptable. That was a gap in the analysis rather than a judgment: averaging a category with 116 pages hides the tail, and the tail is where the recoverable demo requests are. Section 04 now lets you expand any category to see the same thing for yourself.

Beaconpath asked us to start tracking what AI assistants say about Gridwell Automations, because 720 people a month search for it and nobody is monitoring the answers. That request is about measurement. When we looked up the same pages in the warehouse we found a second reason to act on it, one Beaconpath cannot see: assistants are already sending a growing number of people to these pages, and those people are not requesting a demo. Visits rose from 174 to 230 over the last four weeks, up 32 percent, and produced 23 demo requests. That is a 10.0 percent conversion rate against 23.4 percent for AI traffic across the site. The pattern inside the group is consistent. The marketing page at /platform/automations converts 16 percent. The help center pages that explain how automations work convert between 0 and 4 percent, and several of the largest convert nobody at all.

1. Start with support.gridwell.example/docs/getting-started-with-automations. It takes 24 visits and converts one of them. The article teaches someone how to build an automation and then ends. Add a button that opens the product with a new automation started, so the reader can do the thing they just read about.
2. Do the same on the four pages that currently convert nobody: ai-workflow-automation-tools, save-time-with-automations, at-a-scheduled-time-automation-trigger and gridwell-automation-walkthroughs. Together they are 25 visits and zero demo requests.
3. Copy what /platform/automations does. It converts 16 percent on the largest share of this traffic, so it is the working example inside the same group.
4. Separately, add the tracking prompts Beaconpath asked for. That is the original request and it is worth doing, but it measures the problem rather than fixing it.

Beaconpath surfaced this theme on search volume alone. The conversion case comes from demo request submissions in the warehouse over the four weeks to 23 August, compared against the four weeks before. Section 08 shows the full check, including the two Beaconpath items we did not adopt.

Traffic is rising and capture is not following, which is the trend route described in section 03. A page group can look acceptable on its average while the newest traffic lands on the weakest pages in it.

Same situation as the previous action, on a different product area, and growing faster. Beaconpath asked for tracking prompts on View Builder because 320 people a month search for it and we have no visibility into what assistants tell them. The warehouse shows that assistants are already acting: visits to these pages went from 146 to 252 in four weeks, up 73 percent, the fastest growth of any theme in this report. Those 252 visits produced 30 demo requests, a rate of 11.9 percent. The two pages built to sell the feature convert respectably, 16.3 percent on the platform page and 20 percent on the main guide. The traffic growth is not landing on those. It is landing on view-builder-record-review, which took 14 visits and converted nobody, and on the permissions documentation, which took 15 visits across two versions and also converted nobody.

1. Fix guides/collaborate/view-builder-record-review first. It is new traffic, 14 visits in four weeks against 3 before, and it has converted nobody in either period. Nobody built that page expecting visitors, so it has no way for a reader to start.
2. Add the same button to support.gridwell.example/docs/getting-started-with-view-builder. At 35 visits it is the second largest page here and converts 8.6 percent.
3. Decide what the permissions pages should do. Two versions of view-builder-permissions took 15 visits between them and converted nobody. Someone reading about permissions may be an existing admin, in which case no demo request is expected and these should be excluded from the group rather than counted against it.
4. Consolidate the duplicate help center paths. docs/getting-started-with-view-builder and v1/docs/getting-started-with-view-builder are the same article at two addresses, splitting 42 visits between them.

Demo request submissions in the warehouse for the four weeks to 23 August against the four before. Beaconpath independently flagged this theme on search volume, with no knowledge of the conversion gap. The two findings arrived from opposite directions and point at the same pages, which is the strongest signal in this report.

This is the clearest example of the mix shift described in section 03. The pages themselves did not get worse. Assistants changed which ones they cite, and the new favourites were never built to convert.

Search Google for "gridwell mcp interfaces automations" and four different Gridwell pages come back on the first page. The help center article is second, two Gridwell community forum threads are third and sixth, and the marketing article written specifically for this topic is fifth. They are competing against each other for the same searcher. On the more common search "gridwell mcp server" the marketing article does not appear anywhere in the first thirty results, while the help center article takes the top spot. When several of your own pages target one search, Google and the AI assistants choose which to show, and here they are choosing the page that offers no way to request a demo.

Conversion data for the page currently winning these searches:

1. Decide which page should win which search. Recommended: the help center article answers "how do I set up the MCP server", the marketing article answers "what can the MCP server do". Right now both are written for the same phrasing.
2. Change the marketing article's title and headings so they stop matching the help center article. Google is currently quoting the help center article using the automations wording the marketing article was written for.
3. Link the two pages to each other. Neither currently does, so a reader who lands on the wrong one has no way to reach the right one.
4. Raise the two community forum threads with whoever runs the community. One is a feature request describing something the product cannot do, and it ranks third for a search where people are choosing a tool.
5. The help center article is winning these searches and converts at 4.5 percent, so give it the action 1 button whatever else is decided.

Live Google results pulled on 26 August 2026, United States. Verified by looking at the actual results page rather than through a keyword tool, because the standard endpoint returns only the best ranking page per search and cannot show two of your own pages competing.

The same pattern appears on "gridwell automations", where six Gridwell pages occupy the first results page and the AI answer quotes five help center articles while ignoring /platform/automations at position one. See section 06.

When someone asks an AI assistant how to find a Gridwell consultant, they are sent to one of two pages, and neither works. Over the eight weeks measured, 87 people arrived across both and none requested a demo. The services directory recorded every visit as a single page view with zero seconds of activity. Google's AI answer compounds the problem by describing gridwell.example/partners as the Services Partner Directory, when that page is actually the application form for consultants who want to join the program.

1. Move the "find a partner" section of /partners above the "apply to become a partner" section. Today the page title, main heading and primary button all speak to consultants applying to join, not to customers looking to hire one.
2. Find out why ecosystem.gridwell.example/consultants shows every visit ending immediately with zero seconds recorded. That combination normally means broken tracking or a page that fails to load, not that people disliked it.
3. The partner directory investigation already covers these pages and found ten broken redirects and no sitemap. Coordinate with that work rather than fixing the same thing twice.
4. Do not judge this on demo requests. Someone hiring an implementation partner is not going to book a product demo, so success here is a referral to a partner, which this data does not measure.

Account creation events from the warehouse, 30 June to 25 August 2026, plus the live Google AI answer captured in section 06.

Included despite low volume because it is a clear defect rather than an optimization, and because every visit ending instantly with zero seconds recorded suggests the true figure may be understated by a measurement fault.

People who arrive from an AI assistant request a demo 23.4 percent of the time. Visitors from every other source combined do so about 4.5 percent of the time. That means each extra 1,000 people an assistant sends is worth roughly 234 new demo requests. Fixing every underperforming page listed above would recover around 125 demo requests in total, so attracting more of this traffic is worth more than improving how the existing traffic converts. The limit is that this channel is still small, around 24,000 visits over eight weeks against millions across the site.

1. Start with /guides/start/what-is-gridwell. It converts at 30.2 percent, better than any page here except the demo request form itself, but only receives 53 visits while "what is gridwell" carries the highest AI demand in the data. The page works, it is just not being found.
2. Rewrite the top of the page as a direct one paragraph answer to "what is Gridwell", then break the rest into short sections that each answer one follow up question. AI assistants quote pages built that way, which is what section 06 shows happening on the automations searches.
3. Beaconpath records no citations for this page. Look at how /articles/best-ai-agent-builders is written, since it has the highest citation rate on the site at 1.18 percent, and copy its structure.
4. Leave /pricing alone. It already converts at 28.1 percent on 1,454 visits.
5. Measure this by how often AI assistants cite the page, not by its conversion rate. The page already converts well, so the goal is being quoted more often.

Account creation events from the warehouse for the rate, DataForSEO AI Optimization keyword data for demand, Beaconpath category 1388 for citations.

AI demand is a modeled estimate built mainly from Google's People Also Ask data, not observed assistant traffic. Values can be compared against each other within that column and must never be expressed as a ratio against Google search volume.