Where I Would Spend on the Worth-It Question

I would take the page that answers "is it worth it" out of the AI-citation budget and leave it in the sales budget, and I would move citation spend toward the comparison question, where vendor-owned domains have their best showing. The page can still convert a person who already found us. On the question of whether the thing is worth the money, this release says engines go to a forum.

The dataset I would re-read before moving the money is the Machine Relations Index Report, September 18, 2026. That page is the deeper dataset. This essay is how I would use it as a founder: which question shape I would fund, which one I would stop funding as a citation asset, and what would put the money back.

The part I would concede first

Vendor-owned domains earn citations. In this release they take 6.9 to 13.8 percent of the top-ten cited slots, depending on the question. They peak on head-to-head comparisons. IBM, Databricks, and Palo Alto Networks hold comparison slots in their own categories. Specification pages, integration docs, and pricing tables get read and cited.

So the website works, on one shape of question. I would keep building those pages. I would stop describing that result as "AI visibility" in the singular.

The Index splits buyer demand into six shapes: best tools in a category, how buyers choose, is it worth it, problem-first research, ranked top lists, and head-to-head comparisons. Same buyers, same categories, a 125-day window, different citation winners. I pulled the top-cited domain in every published segment and counted the class of source that held it.

Question shape	Community	Vendor's own site	Editorial	Other classes
Is it worth it	7 of 13	1 of 13	0 of 13	5 of 13
Problem-first research	7 of 13	0 of 13	0 of 13	6 of 13
Best tools	3 of 13	2 of 13	4 of 13	4 of 13
How buyers choose	2 of 13	1 of 13	2 of 13	8 of 13
Head-to-head comparison	1 of 13	1 of 13	0 of 13	11 of 13
Ranked top lists	1 of 13	1 of 13	3 of 13	8 of 13

A vendor's own domain wins the single top slot in at most 2 of 13 segments in any shape, and in comparison, the shape friendliest to vendors, in 1. The shape where vendor domains take the largest share of the full top ten is comparison, at 13.8 percent, and even there they take the single best slot once. I would rather own a shape where our domain can hold a slot than rent a lower slot on the shape where it cannot.

The question that sits next to the money

"Is it worth it" is the last question before a buyer spends. Across the 13 categories that publish a rate for that shape, reddit.com holds the top cited slot in 7. One vendor-owned domain holds it. These are the winners.

Category	Top cited domain	Class	Citation rate
AI infrastructure	reddit.com	Community	32.1%
Consumer finance	reddit.com	Community	31.8%
Consumer products	reddit.com	Community	27.9%
AI visibility and GEO	reddit.com	Community	27.7%
Emergent prosumer	reddit.com	Community	25.0%
Education and training	reddit.com	Community	23.0%
Family software	reddit.com	Community	21.0%
Consumer health	uclahealth.org	Unclassified	37.7%
Deep-tech hardware	vlsishuttle.com	Unclassified	35.2%
Cybersecurity	paloaltonetworks.com	Vendor-owned	28.2%
Enterprise software	erpresearch.com	Market database	24.3%
Fintech	paymentsandrisk.com	Unclassified	23.2%
AI security and privacy	witness.ai	Unclassified	21.5%

Seven of thirteen. In seven of those thirteen categories, no vendor-owned domain appears anywhere in the top ten for that question. Across all 13 segments there are 130 top-ten slots, and vendor-owned domains hold 12 of them. The worth question was observed 1,522 times in this window. The engines answering it went to a forum, a health system, a niche hardware site, a security vendor, a market database, or an unclassified domain. They rarely went to the seller.

The words I would use for a question shape, a citation rate, and a source class are in the glossary. The work of earning a citation the seller's own page cannot win alone is the discipline on Machine Relations.

Buyers said this before the engines did

TrustRadius, in its 2026 Buying Disconnect report, found that 63 percent of buyers used AI during their purchase, 94 percent of those fact-check what it tells them, and 74 percent still use reviews to decide. HG Insights summarized the same study as trust, then verify.

SurveyMonkey and Reddit published research in March 2026 that puts peer recommendations at 73 percent trust, ahead of vendor websites at 55 percent, search engines at 54 percent, review sites at 46 percent, and AI chatbots at 39 percent. Reddit co-sponsored that study, so I weight it as an interested party. I still weight it, because it agrees with an index Reddit did not run.

G2's 2026 buyer behavior research found review sites at 38 percent overtaking AI chatbots at 37 percent as the source shaping a shortlist, with finance involvement in software decisions moving from 31 percent to 46 percent in a year.

Buyers reported more trust in peers than in vendors. The engines that read the web for them cite peers over vendors on the question where that trust matters. Two methods, one pattern. I would not call it causation. I would call it enough to stop paying twice for one page.

The exception is the instruction

Cybersecurity breaks the pattern, and it is the most useful row in the release. Five of the ten most-cited domains for "is it worth it" are vendor-owned, and a vendor holds the top slot.

Domain	Class	Citation rate
paloaltonetworks.com	Vendor-owned	28.2%
cyware.com	Unclassified	25.2%
reddit.com	Community	19.1%
swimlane.com	Unclassified	15.3%
sentinelone.com	Vendor-owned	14.5%
breachsense.com	Market database	13.0%
microsoft.com	Vendor-owned	13.0%
rapid7.com	Vendor-owned	13.0%
techtarget.com	Editorial	13.0%
ibm.com	Vendor-owned	11.5%

Those five vendors are not uniformly large. Microsoft sits at rank 12 of 22,179 cited domains, confidence A, with citations in 13 categories. Rapid7 sits at rank 467, confidence C, citations in 2 categories. Rapid7 is cited in 31 of 15,782 observed runs across the whole index, and it still holds a top-ten slot on the cybersecurity worth question. The segment has 131 observed runs and 7 distinct run dates, so it clears the same floor as every other published segment.

Overall authority is not what put them there. Concentration is. Each of the five is cited inside cybersecurity and the adjacent security category, and the two smallest publish almost nowhere else. A founder in cybersecurity and a founder in family software are reading two different markets. I would open my own category's segment, the way I would open cybersecurity's worth answers, and count vendor domains before I copied anyone's playbook.

The other shapes, so the worth row does not eat the budget

The worth row is the one next to the money, and it is a bad template for the other five shapes. I would read them before I moved every content dollar into comparison just because comparison is where vendors look least hopeless.

Problem-first research looks like the worth question from the seller's side of the table: community sources hold the top slot in 7 of 13 segments, a vendor's own site holds it in 0 of 13, editorial in 0 of 13, and other classes in 6 of 13. If a buyer is still naming the problem, this release says the engine is not starting on the vendor's domain. I would not fund a "what is the problem" page as a citation asset and expect it to behave like a docs page.

Best tools is the shape where editorial actually shows up at the top: community 3 of 13, vendor 2 of 13, editorial 4 of 13, other 4 of 13. A listicle query is closer to a publication strategy than a forum strategy. I would not copy the Reddit plan from the worth row into a "best tools" query and call it consistency.

How buyers choose is mostly the residual class: community 2, vendor 1, editorial 2, other 8 of 13. The HR buyer-choice work I have done elsewhere sits in this shape, and the lesson I would carry back is that "other" is doing the winning. I would open the segment and name the domains before I assumed they were vendors.

Head-to-head comparison is where I would move citation money, and the table says why in two different units. Vendor domains take the single top slot only once, same as several other shapes, but they take 13.8 percent of all top-ten slots, the high end of the 6.9 to 13.8 percent range and roughly double their share on best tools. Eleven of 13 top slots in this shape are "other," so even the friendliest shape is not a vendor sweep. It is the shape where a specification page, an integration doc, or a pricing table has the best chance of being used as evidence. IBM, Databricks, and Palo Alto Networks holding comparison slots in their categories is the existence proof. I would build the comparison page to be cited, and I would judge it on whether it enters a top ten, not on whether it becomes the forum.

Ranked top lists give editorial 3 of 13 top slots, community 1, vendor 1, other 8. That is a publication shape again, closer to best tools than to worth. I would not brief one "AI visibility program" across all six. The release is explicit that they behave like separate markets over the same 125 days.

The worth exception remains cybersecurity, and I would not let it leak. Five of ten domains vendor-owned, Palo Alto at 28.2 percent, Rapid7 at rank 467 overall and still in that top ten on 31 of 15,782 runs. Family software, at the other extreme, is a Reddit win at 21.0 percent with the vendor absent from the kind of top ten the seven zero-vendor categories describe. A founder who copies the security page onto a consumer category is copying the wrong market. I would open the category segment, count vendor domains, and only then decide whether the worth page stays in the citation budget. The default from this release, outside a concentrated exception, is that it does not.

The decision

Four moves, one budget.

I would leave the worth page up for people who already found us, and I would remove it from the citation budget. In a category where no vendor domain holds a top-ten slot on that question, paying to make machines cite it is paying for a job the release says they do not give the seller. The page can stay a sales asset.

I would read my category before I generalized. Seven of thirteen categories have zero vendor presence in the worth answer. Cybersecurity has five of ten. The playbook that is rational in one is busywork in the other.

I would move citation money to the comparison shape, where vendor-owned domains take 13.8 percent of top-ten slots, roughly double their share on the best-tools shape. I would rather hold the question our domain can win.

For the worth question itself, I would stop publishing our way into the answer and start earning the citations the engines already prefer. That means customers stating specific numbers in public. It does not mean a company account posting in a forum, which gets the thread removed and deserves to. It is a product and support problem before it is a marketing one, and it is slower than a rewrite.

Field	What I would write down
Decision	Remove the worth page from the citation budget and reallocate to the comparison shape
Date	2026-09-18
Evidence	MRI release mri_score_v2.0, generated 2026-09-18, window 2026-05-10 to 2026-09-18
Scope	13 published "is it worth it" segments, 1,522 observed runs
Reversal	A release where vendor-owned domains hold the top slot in 4 or more of the 13 worth segments, or where our own category shows a vendor domain in the top ten
Review	The next quarterly release read

Scope is those 13 published segments and the comparison shape I would fund instead. I would not extend the decision to best-tools or to news, where the table already shows editorial and community sources with different shares. The machine-readable release is the file I would re-filter, not a screenshot of this essay.

What would change my mind is the reversal line. If vendor-owned domains hold the top slot in 4 or more of the 13 worth segments, the "engines will not cite the seller here" conclusion has weakened and I would put the page back into the citation budget. If our own category shows a vendor domain in the top ten on that question, I would treat us like the cybersecurity exception and fund the vendor page, narrowly, the way Rapid7's 31 runs across the whole index still coincide with a top-ten slot inside one concentrated category. A single flattering run would not meet either test.

What the release is

The index observes six engines: ChatGPT, Claude, Gemini, Google AI Mode, Google AI Overviews, and Perplexity. This release covers 15,782 answer runs, 124,397 citation events, and 22,179 cited domains over 125 days. A segment publishes a rate only after 10 observed runs across 7 distinct dates. The index runs the market's questions, not questions about one brand, so it cannot tell me whether my pages are performing. Citation here is correlation. I would not publish a sentence that says a page caused a rate.

Reddit was cited by four of the six engines in this window, not all six. Google's crawler documentation describes Google-Extended as the control for Gemini grounding, separate from Search inclusion, and Google's AI features guidance describes those surfaces drawing on the ordinary index. Perplexity documents query-time search over live sources. Each engine assembles its own answer. The domain that dominates the worth question in four engines is absent from the other two. I would not brief the company as if all six engines had agreed.

FAQ

What is a question shape?

A pattern the Index uses to group buyer queries. There are six: best tools, how buyers choose, is it worth it, problem-first research, ranked top lists, and head-to-head comparisons. Each is measured per category. I would not average them into one visibility score.

Does the company site stop mattering?

Vendor-owned domains hold 13.8 percent of top-ten slots in comparison answers, their strongest shape, and 6.9 to 13.8 percent depending on the question. The site matters on some questions. On the worth question it holds the top slot in 1 of 13 segments and 12 of 130 top-ten slots. I would fund it where that share is real.

Why does Reddit win so often?

The index reports what was cited. It does not report a motive. The pattern lines up with the March 2026 trust figures, peer recommendations at 73 percent and vendor websites at 55 percent. Consistent, and not a causal proof. Reddit's own sponsorship of that survey is a reason to keep the index, which Reddit did not run, as the deciding file.

Is the cybersecurity exception noise?

Five of ten top-cited domains in that segment are vendor-owned, against a pattern of zero to two elsewhere, over 131 observed runs and 7 dates. It clears the same floor as every published segment. Microsoft's rank 12 of 22,179 and Rapid7's rank 467 can share a top ten because the segment rewards concentration, not global fame.

How would I reproduce the count?

Filter the public release to published strata with question shape is_x_worth, rank each category by citation rate, and join the source role from the domain record. The release id is mri_score_v2.0, generated 2026-09-18. If my filter disagrees with the table above, the filter wins and this decision gets rewritten.