New researchWhat do the best MSP websites do differently?What do the best MSP websites do differently?See the research
What do the best MSP websites do differently? See the research
Try for free
MSP Website Benchmark

Methodology: MSP Website Benchmark 2026

By Michael Smith · Published

This page explains how we ran We scored 50 MSP websites: who we picked, what we checked, how the score works and what the data can't tell you.

The sample

We studied 50 managed IT service providers (MSPs) in 10 UK towns: London, Birmingham, Manchester, Leeds, Glasgow, Cardiff, Belfast, Norwich, Exeter and Sunderland. Five firms come from each town.

  • Visible (25 firms): on Google's first page for IT support in their town when we drew the sample.
  • Less visible (25 firms): not on the first page for those searches.

Candidates came from Google search and Maps results and from Companies House. To be included, a firm had to be UK-headquartered and independent, with managed IT as its main service and a website at least a year old. Firms were then drawn at random within each town and group. No firm was hand-picked, and no NetNav customer, partner or agency client was included.

Two Belfast websites showed a maintenance page or blocked our crawler during collection, so findings about websites are out of 48. Findings about Google and AI cover all 50.

What we collected

Every firm has 227 data points, collected on 9 November 2026. All checks were passive:

  • public pages only, at no more than one page a second, identified as NetNav's crawler (about the crawler)
  • no forms submitted, no calls booked, no logins, no admin or hidden URLs tried
  • no vulnerability scanning, port scanning or fuzzing
  • staging or test hostnames seen in public certificate logs were noted, never visited

Sources: Google search results (mobile, set to each town), Google Business Profiles, the AI tools below, the websites themselves (pages, headers, screenshots and page code), DNS records, Lighthouse lab tests on a mid-range phone, automated accessibility checks and a cookie check from a fresh browser.

How we asked AI

We asked ChatGPT, Gemini and Google's AI Overview six questions a buyer would ask, for each town:

  1. Who are the best IT support companies in [town]?
  2. Recommend a managed IT service provider for a 30-person business in [town]
  3. Which MSPs in [town] are good for law firms?
  4. I need a new IT support company in [town]. Who should I talk to?
  5. Who offers cyber security and IT support for small businesses in [town]?
  6. What are the top-rated IT companies near [town] for Microsoft 365 support?

Each question was asked three times, in fresh sessions, set to the UK where the tool allows it. A firm counts as recommended only if it was named in at least two of the three runs. We also asked each tool about each firm by name and checked what it said. In total we analysed 690 AI answers. AI answers change over time, so these results describe November 2026.

How judgement fields were read

Some checks need judgement: does the homepage name a sector, is there a case study with a measurable result, what kind of button leads the page. One AI model read each site's screenshots and page text for these and quoted the evidence for every answer. We then re-checked every borderline answer against that evidence and corrected 25 of them.

Stock phrases are counted from a fixed list: trusted IT partner, peace of mind, tailored solutions, one-stop shop, cutting-edge, end-to-end, seamless, technology partner, focus on your business, we take care of IT, IT that just works, proactive approach, best-in-class, world-class, solutions provider, your IT department.

The MSP Website Score

The score is out of 100, across six pillars. Found and Recommended measure how a firm shows up on Google and in AI answers. The other four pillars (70 points) measure the website itself. We call these four "website quality".

A check we couldn't measure scores 0, with two exceptions. A CMS version that isn't on show counts as not outdated. A site with no cookie banner gets the "Reject all" point if it loads no trackers.

We tested the weights by raising and lowering each pillar by half. At most two firms moved in or out of the top 10, so we publish the single score.

Found (20 points)

Field What we check Points How it scores
rank_brand Position for the firm's own name 2 1 = 2 pts; 2-3 = 1; else 0
rank_service_place_1 Best position for "IT support [place]" 3 Top 3 = 3; 4-10 = 2; 11-20 = 1; else 0
rank_service_place_2 Best position for "managed IT services [place]" 3 Top 3 = 3; 4-10 = 2; 11-20 = 1; else 0
rank_service_place_3 Best position for "IT support company [place]" 1 Top 10 = 1; else 0
rank_service_place_4 Best position for "cyber security [place]" 1 Top 10 = 1; else 0
map_pack_any In the map pack for any of the 4 queries 2 Yes = 2
ranked_keywords_count Number of keywords the domain ranks for in top 100 (UK) 1 Median or above = 1
non_brand_top10_count Keywords in top 10 that don't contain the brand name 1 >= 5 = 1
gbp_found Business Profile found and matches the firm 1 Yes = 1
gbp_primary_category Primary category 1 Category contains "computer support" or "IT" = 1
gbp_review_count Number of reviews 2 >= 25 = 1; >= 75 = 2
gbp_newest_review_days Days since newest review 1 <= 60 days = 1
gbp_owner_replies_pct Share of last 10 reviews with an owner reply 1 >= 50% = 1

The two checks add up to 9 raw points, scaled to 10.

Field What we check Points How it scores
ai_prompts_named Prompts (of 6 per town) where the firm is named in at least 2 of 3 runs, in any AI tool 6 1 per prompt, max 6
ai_engines_named AI tools (of 3) that name the firm at least once 3 1 per tool, max 3

Clear (20 points)

Field What we check Points How it scores
stock_phrase_count Stock phrases on the homepage (list below) 2 0-1 = 2; 2-3 = 1; 4+ = 0
sameness_score Similarity of the homepage's opening copy to the other sites 2 Least similar third = 2; middle third = 1
sector_focus Homepage names a sector or client type it specialises in 2 headline = 2; secondary = 1
audience_size_stated States the size of business it serves 1 Yes = 1
outcome_in_headline Headline states an outcome, not just a service 1 Yes = 1
pricing_published How pricing is shown 3 exact or from_price = 3; model_only = 2; none = 0
response_time_stated Response or resolution times published 2 Yes = 2
contract_terms_stated Contract length or notice period stated 1 Yes = 1
onboarding_described Describes how onboarding or switching provider works 2 Yes = 2
service_pages_count Separate pages for distinct services 1 >= 4 = 1
posts_last_12m Posts published in the last 12 months 2 >= 6 = 2; 1-5 = 1
author_named Posts carry a named author 1 Yes = 1

Trusted (15 points)

Field What we check Points How it scores
accred_cyber_essentials Cyber Essentials or Cyber Essentials Plus shown 2 ce or ce_plus = 2
accred_iso27001 ISO 27001 shown 1 Yes = 1
accred_vendor Vendor partner status shown (e.g. Microsoft Solutions Partner) 1 Any = 1
case_studies_count Case studies published 2 >= 3 = 2; 1-2 = 1
case_studies_with_results Case studies with a measurable result 3 >= 1 = 3
testimonials_named Testimonials with a person's name and company 2 >= 3 = 2; 1-2 = 1
reviews_on_site Third-party reviews or rating shown on site 1 Yes = 1
team_real_photos Team page with real photos and names 2 Yes = 2
client_logos Client logos shown 1 Yes = 1

Easy to enquire (15 points)

Field What we check Points How it scores
cta_type Type of main call to action 3 book_call or free_assessment = 3; quote = 2; generic = 0
cta_above_fold_mobile Main CTA visible on first mobile screen (390x844) 2 Yes = 2
phone_in_header Phone number in header 1 Yes = 1
click_to_call Phone is a tel: link on mobile 1 Yes = 1
form_fields_total Fields on the main contact form 3 <= 5 = 3; 6-8 = 1; 9+ = 0
reply_time_promised States how quickly enquiries get a reply 2 Yes = 2
booking_link Online booking for a call 1 Yes = 1
free_offer Free assessment, audit or review offered 2 Yes = 2

Sound (20 points)

Field What we check Points How it scores
https_valid Loads over HTTPS with a valid certificate; http redirects to https 1 Yes = 1
hsts Strict-Transport-Security header present 1 Yes = 1
csp Content-Security-Policy header present 1 Yes = 1
frame_protection X-Frame-Options or CSP frame-ancestors present 0.5 Yes = 0.5
x_content_type_options X-Content-Type-Options: nosniff present 0.5 Yes = 0.5
tls_min_version Lowest TLS version accepted 1 1.2 or higher = 1
cms_version_exposed CMS version visible in HTML or headers 1 No = 1
cms_outdated Exposed CMS version is behind the latest release 1 Not outdated = 1 (a hidden version counts as not outdated)
spf SPF record present for the domain 0.5 Yes = 0.5
dmarc_policy DMARC policy 1 quarantine/reject = 1; none = 0.5
mixed_content Mixed content on the 4 key pages 0.5 No = 0.5
cwv_pass Passes Core Web Vitals on mobile 2 Yes = 2; where Google has no real-user data, lab LCP of 2.5s or less = 1
lighthouse_perf_mobile Lighthouse performance score, mobile, homepage 1 >= 70 = 1
indexing_blocks Key pages blocked by noindex or robots.txt 1 0 = 1
broken_internal_links Broken internal links found 1 0 = 1; 1-5 = 0.5
title_meta_issues Key pages with missing/duplicate title or meta description 0.5 0 = 0.5
schema_org Organization or LocalBusiness structured data present 0.5 Yes = 0.5
a11y_issues_serious Serious/critical automated WCAG issues on 3 key pages 1.5 0-2 = 1.5; 3-10 = 0.5
a11y_form_labels Contact form fields all have labels 0.5 Yes = 0.5
trackers_before_consent Non-essential trackers firing before any consent choice 1.5 0 = 1.5
reject_first_layer "Reject all" on the cookie banner's first layer 0.5 Yes = 0.5 (no banner needed, and the point given, when nothing tracks)
privacy_policy Privacy policy linked from footer 0.5 Yes = 0.5
company_details_shown Registered name, number and office shown 0.5 Yes = 0.5

How we report

  • Counts first. Every figure is a count of the group, such as "27 of 48", never a percentage on its own.
  • Ranges. With 50 firms, every count has a range. The table below gives the 95% range for similar firms (Wilson interval).
  • Group differences. Differences between the visible and less visible groups are tested with Fisher's exact test, and only reported as a difference at p < 0.05. Each one is re-checked among small firms only, and with the groups re-drawn from the rankings on the day of the scan.
  • No circular findings. The groups are defined by Google ranking, so ranking is never reported as a difference between them.
  • No cause and effect. We say things "go together", never that one causes another.
  • Medians, not averages, for scores.
  • No sub-groups under 10 firms, and no results for a single town.
Finding Count 95% range for similar firms
2+ stock phrases (Q3) 27 of 48 42%-69%
4+ stock phrases (Q3) 15 of 48 20%-45%
outcome in headline (Q3) 20 of 48 29%-56%
sector in headline (Q3) 3 of 48 2%-17%
sector named anywhere (Q3) 13 of 48 17%-41%
any pricing (Q4) 21 of 48 31%-58%
a figure (exact/from) (Q4) 4 of 48 3%-20%
response time stated (Q4) 19 of 48 27%-54%
contract terms stated (Q4) 7 of 48 7%-27%
onboarding described (Q4) 19 of 48 27%-54%
Cyber Essentials (any) (Q5) 26 of 48 40%-67%
case study (Q5) 20 of 48 29%-56%
case study with a result (Q5) 8 of 48 9%-30%
named testimonial (Q5) 23 of 48 34%-62%
real team photos (Q5) 21 of 48 31%-58%
generic CTA (contact/other/none) (Q6) 25 of 48 38%-66%
booking link (Q6) 8 of 48 9%-30%
reply time promised (Q6) 9 of 48 10%-32%
no form found (Q6) 9 of 48 10%-32%
DMARC enforced (all 50) (Q7) 32 of 50 50%-76%
HSTS (Q7) 22 of 48 33%-60%
CSP (Q7) 9 of 48 10%-32%
CMS version on show (Q7) 26 of 48 40%-67%
plugin/library with known issues (Q7) 19 of 48 27%-54%
Lighthouse mobile 70+ (Q8) 26 of 48 40%-67%
3+ serious accessibility issues (Q8) 27 of 48 42%-69%
trackers before consent (Q8) 31 of 48 50%-77%
reject on first layer (Q8) 16 of 48 22%-47%
no cookie banner (Q8) 22 of 48 33%-60%
no banner AND trackers (Q8) 12 of 48 15%-39%
no organic link for own name (all 50) (Q1) 23 of 50 33%-60%
GBP found (all 50) (Q1) 36 of 50 58%-83%
25+ reviews (all 50) (Q1) 8 of 50 8%-29%
4+ service pages (Q1) 28 of 48 44%-71%

What this study can't tell you

50 firms can't describe every UK MSP. The sample is one point in time, AI answers change from week to week, and a website can change the day after we look at it. We measure what websites show, not how good a firm's service is.

Firms in the study

We don't name any firm without its written permission. Firms with a security issue on their own website are told privately; we never publish anything that identifies a firm with a weakness. Any firm can ask to see its own results, correct them or be removed: email [email protected].

Extended scan (12 to 13 November 2026)

Two research pieces, SEO for MSPs and MSP website security, also use a larger scan of MSP websites. The report above doesn't use it, and its figures are unchanged.

Sample. The benchmark's screening found 276 UK managed IT providers with a readable website in the same 10 towns (102 visible on Google, 174 less visible). We drew 250 of them at random, keeping that split. One sent our crawler a bot check and was left out, so the scan covers 249 firms: 90 visible and 159 less visible. 43 of the benchmark's 50 firms are in it. Our crawler couldn't read 8 of the websites, so website checks are out of 241 and email and domain checks are out of 249.

What it checks. The same collectors as the benchmark, run on 12 to 13 November 2026: security headers, HTTPS, email security (SPF, DKIM, DMARC), software versions and known vulnerabilities, cookies before consent, speed (lab tests on a mid-range phone), accessibility, pages and content. Some fields, such as location pages and accreditations shown, are judged by AI from the page text and screenshots. Two runs of the same model agreed on about 7 in 10 of those answers, so we report them as broad patterns only.

What it doesn't check. No new Google rankings, Business Profiles, reviews, backlinks or AI answers. Where the pieces mention those, the figures come from the benchmark's searches on 9 to 10 November 2026.

Groups and tests. "Visible" means on Google's first page for IT support in the firm's town when the sample was drawn. Differences between the groups are tested with Fisher's exact test and reported only at p < 0.05, as in the report.

Security. All checks are passive: what any visitor's browser or a public lookup can see. We don't name firms. Anything serious is shared privately with the firm concerned.

Sources

Free audit, ready in 20 minutes

See what's holding your website back.

Enter your website. You'll get your score, your plan and the first things to fix.

Try for free
No expertise needed
Your Blueprint in 20 minutes
No credit card required