We fix technical SEO, add structured data and write answer-first content, so Google can rank your pages and AI answers can quote them. We report AI citations beside rankings.
best payroll software for mid-size companies in IndiaSearch
Results pageOrganic · 10 blue links
01reviewsite-b.exampleReview site B › Buyer guidesPayroll software compared: 9 platforms for 200–2,000 staffSide-by-side comparison of pricing, statutory coverage and support for growing Indian teams. Updated this month.
02yourcompany.comYour company › Payroll › Mid-sizePayroll for mid-size companies in IndiaMulti-state payroll with EPF, ESI, PT and TDS filed on time, an audit trail per run, and a finance-owned approval step before money moves.SitelinksPricingPer employee per month, and what changes the band›Statutory coverageEPF, ESI, professional tax and TDS, state by state›ImplementationSix to twelve weeks, with a parallel run›
03tradepublication-a.exampleTrade publication A › Finance opsWhat finance teams get wrong when they change payrollSix failure patterns from mid-market payroll migrations, and the checks that catch them before the first live run.
People also ask
What is the best payroll software for 500 employees in India?+How much does payroll software cost per employee per month?+Which payroll platforms handle multi-state statutory filings?+
Generated answerThree sources cited
Illustrative generated answers for the same query. Hosts, wording and citations are fictional. Choose an engine to read its answer, the sources it cited and what produced the mention.
Google · generated answer above the results
For teams of 200–2,000 in India, buyers shortlist on statutory coverage, multi-state payroll and a clean audit trail.
Most comparisons weigh EPF, ESI, professional tax and TDS handling first1, then look at how approvals and evidence work before money moves2. Implementation is usually quoted in weeks, not days, because payroll history has to be migrated and reconciled3.
Sources
1reviewsite-b.example
2yourcompany.com
3tradepublication-a.example
What moved this answer
Page/payroll/mid-size-india — answers in the first fifty words
EntityOrganization @id resolves; the profiles state the same facts
Citationreviewsite-b.example names you in its buyer guide
Assistant answer with linked sources
Shortlist on compliance depth, then on how the platform proves what it did.
Buyer guides for this size band consistently rank multi-state statutory filing above price1. Vendors that publish a per-run audit trail and a named approval step tend to clear finance and internal audit faster2. Expect a parallel run before cutover3.
Sources
1reviewsite-b.example
2yourcompany.com
3forum-c.example
What moved this answer
Page/compare/platforms — a server-rendered comparison table
EntityA named author with a checkable profile on the page
Citationforum-c.example answer written by your own engineer
Answer with numbered citations
Compliance coverage, migration effort and audit evidence decide most mid-market payroll choices.
Sources agree that statutory scope is the first filter1 and that migration of historical payroll data is the main schedule risk2. Pricing is generally quoted per employee per month, with a floor for smaller headcounts3.
EntityResolves, but no page of yours matches this question
CitationNo third-party source names you for this prompt
An illustrative ranked results page, shown for contrast beside the answer panel above. The query “best payroll software for mid-size companies in India” returns a review site first; “Your company” second, with sitelinks to its pricing, statutory coverage and implementation pages; and a trade publication third, followed by three People-also-ask questions. All results, hosts and figures are fictional.
01 · The shiftWhy now
AI answers mean fewer clicks.
Buyers ask the same questions, but AI Overviews, ChatGPT and Perplexity now answer them directly and cite a few sources. A visit is no longer certain.
ThenThe ten-links journey
Query
typed into the box
Ten links
scanned and compared
Click
usually more than one
Your site
the answer is on your page
One query, a page of options, a visit.
NowThe answer-first journey
Query
typed, spoken or prompted
Generated answer
assembled from a few sources
Citation
shown, not always followed
Your site
or no visit at all
One query, an answer, sometimes a visit.
The box looks the same, but it now returns a written answer.
What we still report
Clicks, sessions and conversions, measured as before.
What we add
Share of answer and citation rate across AI answers.
A few sources per answer
A results page lists ten links; an AI answer usually cites a few, so the aim is to be one of them.
Fewer places to appear
Citations without a click
When an AI answer names your brand, the buyer has seen you before any visit, so citations are worth measuring.
Measured and sampled
SEO as the foundation
Google says its AI features need no special markup beyond Search Essentials, so technical SEO remains the groundwork.
Same groundwork
02 · The lensShare of answer
Where AI answers cite you, and where they miss you.
Promptpayroll software for a 500-person companyWhere the answer is read often on a phone, before any clickPhoto: Jonas Leupe / Unsplash
We sample the prompts your buyers use across search and AI engines and show where you appear. Open a cell to see the answer, its sources and the fix.
Share-of-Answer Lens · Your company · payroll, India
Panel of 12 prompts · 4 engines · 20 runs per engine per week
Illustrative share-of-answer panel. Twelve buyer prompts, grouped by intent, against four engines. Each cell records whether the placeholder company was cited with a link, mentioned without a link, or absent from that engine's answer.
Buyer prompt
Google organicOrganicRanked results
Google AI OverviewAI OverviewGenerated summary
ChatGPT searchChatGPTAssistant answer
PerplexityPerplexityCited answer
DiscoverLearning the problem exists
what is multi-state payroll and why is it hard
how do mid-size companies in India run payroll today
payroll compliance checklist for a 500-person company
signs your payroll process has outgrown spreadsheets
CompareDrawing up a shortlist
best payroll software for mid-size companies in India
payroll platforms compared for 200 to 2000 employees
alternatives to an in-house payroll team
payroll software pricing per employee per month
DecideGetting it through the business
how long does a payroll migration take
questions to ask a payroll vendor before signing
which payroll vendors publish audit evidence
payroll software implementation checklist
Citednamed with a link to your page
Mentionednamed without a link
Absentnot named at all
Share of answer by intentSampled estimate
DiscoverLearning the problem exists13%cited
CompareDrawing up a shortlist13%cited
DecideGetting it through the business31%cited
How this is measured. AI answers vary from run to run, so one check proves little. We sample each prompt twenty times per engine weekly, from a clean session in the buyer's market, and report a rate with its sample size. Organic positions come from Search Console where available; AI-answer figures are labelled as sampled estimates. Figures shown here are illustrative.
03 · FoundationsCrawl · render · index
Technical SEO that lets crawlers read your pages.
Rankings and AI citations both start with a crawler fetching your page. We check five stages, because a page can fail at any one without anyone noticing.
01
Discover
Sitemaps · internal links · IndexNow
XML sitemaps under 50,000 URLs and 50 MB each
Every important page reachable in three clicks
IndexNow ping on publish for Bing and partners
02
Crawl
robots.txt · status codes · crawl budget
robots.txt rules checked for each user-agent
No redirect chains: one hop, 301, then stop
Server response under 600 ms, the threshold Lighthouse audits against
03
Render
JavaScript rendering
Primary content present in the server HTML
Nothing important behind a click-only tab
Rendered HTML confirmed in URL inspection
04
Index
canonicals · duplicates · parameters
One canonical per piece of content, self-referencing
Faceted and tracking parameters handled deliberately
noindex only where it is intended
05
Serve
Core Web Vitals · HTTPS · mobile
LCP, INP and CLS rated good at the 75th percentile
/pricing?plan=growth&ref=ad200 · duplicate of /pricing
/resources/checklist.pdf200 · gated, not indexable
/trust/audit-evidenceBlocked by robots.txt
/blog/old-payroll-guide301 → 301 → 200 · chain
/compare/platforms200 · indexed
/careers/payroll-analyst-2021404 · still in the sitemap
An illustrative crawl log for the placeholder company: eight sampled URLs with their status. Two are clean, four carry warnings (client-side rendering, a duplicate parameter URL, a gated PDF and a redirect chain) and two are broken: a trust page blocked in robots.txt and a removed job page still listed in the sitemap.
robots.txt · user-agent rules
GooglebotGoogle Search crawling and indexing
BingbotBing crawling and indexing
OAI-SearchBotFetches pages for ChatGPT search results
GPTBotCrawls content to train OpenAI models
PerplexityBotCrawls to index pages for Perplexity answers
Google-ExtendedControls Gemini use of already-crawled content
The distinction that costs traffic. Blocking all “AI crawlers” in one line also removes you from ChatGPT search and Perplexity citations. Training access and search access are separate decisions with separate agents. We write the rules per agent, record the reason and re-check them after every platform change.
Googlebot · allowed
Allows Google to crawl and index the page, the basis for both organic results and AI Overviews.
Blocking it removes you from Google Search entirely, including AI Overviews.
Core Web VitalsGood thresholds, measured at the 75th percentile of real users over 28 days
LCP≤ 2.5 sLargest Contentful Paintwhen the main content has rendered
INP≤ 200 msInteraction to Next Painthow quickly the page answers a tap
CLS≤ 0.1Cumulative Layout Shifthow much the layout moves while loading
04 · EntitiesStructured data
Clear company facts engines can trust.
Engines cite companies they can identify with confidence. We add structured data and keep your name, address, products and people consistent everywhere they appear.
makesOfferProductWhat you sell, named consistently
employeePersonNamed authors and spokespeople
addressPostalAddressOne address, everywhere the same
reviewReviewRatings you did not write yourself
profilessameAsThe profiles that confirm the entity
authorArticleWho wrote it, and when it changed
mainEntityFAQPageQuestions, answered on the page
itemListElementBreadcrumbListWhere the page sits in the site
yourcompany.com · /#organizationSchema.org
application/ld+json
{"@context": "https://schema.org","@type": "Organization","@id": "https://yourcompany.com/#organization","name": "Your company","url": "https://yourcompany.com/","logo": "https://yourcompany.com/logo.png","description": "Payroll and compliance software for mid-size companies in India.","foundingDate": "2014","address": {"@type": "PostalAddress","addressLocality": "Bengaluru","addressRegion": "KA","addressCountry": "IN"},"contactPoint": [{"@type": "ContactPoint","contactType": "sales","areaServed": "IN"}],"sameAs": ["https://www.linkedin.com/company/yourcompany","https://www.wikidata.org/wiki/Q00000000","https://github.com/yourcompany"]}
An illustrative JSON-LD block for the placeholder company: an Organization with a stable identifier, name, URL, logo, description, founding date, a postal address in Bengaluru, a sales contact point for India and three sameAs profiles.
What markup does and does not do. Valid structured data makes a page eligible for a rich result and helps engines identify your company. It does not guarantee a rich result or rank a page. We mark up only what is true on the page.
Types we implement, and where
OrganizationOnce, sitewide, with a stable @idA consistent entity every engine can resolve
Product / OfferProduct and pricing pagesPrice, availability and rating in the result
FAQPagePages that answer common questionsEligibility only — Google restricted FAQ rich results to authoritative government and health sites in 2023
BreadcrumbListEvery page below the home pageA readable path instead of a raw URL
Company facts that match everywhere
01One legal name
The same registered name on the site, the profiles and the filings. Trading names go in alternateName, never in place of the legal one.
02One address, one number
Identical formatting everywhere, down to the punctuation. Conflicting details are a common reason engines cannot identify a company.
03One canonical URL per entity
A stable @id that never changes, so every later reference resolves to the same thing.
04Claimed profiles
Business profiles, professional networks and developer or industry directories, claimed and kept current, then declared in sameAs.
05 · ContentContent writing
Pages that AI answers can quote.
We write pages that give the answer first and back it up, so readers and AI assistants can find the part worth quoting. Seven parts do most of the work.
01
A direct answer first
The first fifty words answer the question, so readers and AI assistants get it without a preamble.
Easy to quote
02
A comparison table
AI answers favour clear comparisons: options in rows, buyers’ criteria in columns, every cell filled or marked not applicable.
Easy to compare
03
Original data
A survey, benchmark or cost breakdown of your own gives AI answers a reason to cite you that no competitor can copy.
Worth citing
04
A named author with credentials
The writer’s name, role and bio, linked to profiles that confirm them, with Article markup whose author points to that Person.
Shows expertise
05
A visible update note
A date that changes only when the content does, with a short note of what was revised.
Up to date
06
Sources, linked and dated
Statutory rules, official guidance and research, each linked to its primary source with the date it was checked.
Easy to check
07
The next questions buyers ask
Taken from sales calls, support tickets and your prompt set, each answered in two or three sentences below the question.
Payroll platforms for mid-size companies in India, compared
For a company of 200–2,000 staff in India, the platform that fits is the one that files every state’s statutory return without manual work and records who approved each run. Coverage and evidence decide it; price rarely does.
02
An illustrative comparison of three unnamed payroll platforms across five criteria.
Criterion
Platform A
Platform B
Platform C
Multi-state PT slabs
All states
18 states
Manual upload
EPF / ESI filing
Automated
Automated
Export only
Approval before payout
Named approver
Any admin
Named approver
Audit trail per run
Immutable log
90-day log
None
Typical go-live
6–8 weeks
4–6 weeks
10–12 weeks
03Original data · from our own migrations
11 of the last 14 migrations lost time to one thing: historical payroll data that did not reconcile against the last filed return.
Sample: 14 mid-market migrations. Figures illustrative for this mock.
04
Written by a named payroll leadTwelve years in Indian statutory payroll · profile linked · marked up as Article > author > Person
05
Updated 14 days ago — added the FY changes to professional tax slabs in two states and re-checked every filing deadline in the table.
06
Sources
1gov-portal.exampleStatutory filing calendar · checked this month
2tradepublication-a.exampleMid-market payroll survey · published last quarter
How long does a payroll migration take?Six to twelve weeks for most mid-size companies, driven by how clean the historical data is.
Can one platform file for every state?Some can. Check professional tax and labour welfare fund coverage state by state before you sign.
An illustrative comparison page for a placeholder company with seven numbered blocks: a direct answer, a five-row table comparing three unnamed platforms, an original data point from fourteen migrations, a named author, a dated update note, three linked sources and two frequently asked questions. Each number matches the note beside it.
How each page is written
Agents read the prompt set, assemble the brief and check each claim against its source. Editors own the argument, the judgement and the byline.
Agents brief, editors write
An agent compiles the brief: prompts where you are absent, support questions, gaps in competitors’ comparisons and sources to cite.
Claims checked against sources
Before publication, an agent flags any sentence its linked source does not support, and the editor decides what changes.
One named editor per page
A named editor approves each page and answers for it. No byline goes on work that person did not review.
Where this comes from. Google asks for helpful, reliable, people-first content and treats pages mass-produced mainly to game rankings as spam, whether a person or a model wrote them. We use AI to help editors work faster and more accurately, never to fill a site with pages nobody asked for.
06 · AuthorityCitations & mentions
Get mentioned where AI finds answers.
AI answers draw on the same few publications, review sites, communities and reference pages again and again. We work to earn you mentions there, disclosed and never bought.
Month 12Illustrative
Sources linking to you
9
Share of answer12-prompt panel · 4 engines
31%
Brand mentions / week
34
Month 000612
Moves an illustrative twelve-month timeline. As the month rises, more of the nine generic sources link to the placeholder company and the existing links thicken.
Source families · select one to isolate it
Trade publication ATrade press · linked month 01
Review site BReview sites · linked month 02
Community forum CCommunities · linked month 03
Analyst brief DDocs & data · linked month 04
Trade publication BTrade press · linked month 05
Standards body HDocs & data · linked month 06
Review site EReview sites · linked month 07
Q&A community FCommunities · linked month 08
Open dataset GDocs & data · linked month 09
Data-led digital PR
Pitches built on a benchmark, survey or analysis of public filings give journalists a number to quote and a reason to link.
Named experts, on the record
Your engineers and specialists quoted under their own names, with profiles that confirm who they are.
Reviews from every customer
Every customer is asked for a review at the same stage, never only the happy ones; comparison answers often cite review sites.
Community answers
People who do the work answer community questions under their own accounts, without advertising and always disclosing who they work for.
What we will not do. No paid links, link exchanges, private blog networks or guest posts written to place a link. Google’s spam policies treat these as link schemes, and a manual action costs more than the links were worth. Every placement is earned with something worth publishing.
07 · ToolbeltTechnologies we work with
The tools we use, by job.
We group our tools by the job they do, and say what each one can measure and what it cannot see.
01Crawl & audit
Can every page be reached, rendered and indexed?
Full-site crawls of the rendered pages, server logs showing what search crawlers fetched and index coverage from the search engines’ own tools.
Measured: a crawl either reached the URL or it did not
Also in this jobScreaming FrogSitebulbPlaywrightServer log filesBing Webmaster Tools
What it reads
Every indexable URL with its status code, canonical and the HTML after JavaScript has run.
What it cannot see
Whether a page deserves to rank. A crawl proves reachability, never relevance.
What it leaves behind
A crawl export and a ranked fix backlog, each row tied to the URLs it affects.
Technologies we work with
Google Search ConsoleSearch, SEO & analytics
LighthouseSearch, SEO & analytics
PageSpeed InsightsSearch, SEO & analytics
PythonLanguages & runtimes
02Measure
What did visitors do?
Clicks, impressions and position from Search Console; sessions and conversions from analytics; events defined once and named the same everywhere.
Measured: first-party data from your own sites
Also in this jobBing Webmaster ToolsConsent management platform
What it reads
Clicks, impressions and average position by query and page; sessions, engaged sessions and conversions.
What it cannot see
Queries Google withholds for privacy, and anything a consent decision stopped us collecting.
What it leaves behind
A Search Console property with a scheduled bulk export, and an event schema named once.
Technologies we work with
Google Search ConsoleSearch, SEO & analytics
Google AnalyticsSearch, SEO & analytics
Google Tag ManagerSearch, SEO & analytics
PostHogSearch, SEO & analytics
03Research
What are buyers asking, and who answers today?
Demand mapped from query data, sales calls and support tickets, then matched to who ranks and who is cited today. Competitor gaps become the content backlog.
Estimated: third-party keyword volumes come from models
Also in this jobAhrefsYour sales call notesSupport ticket exports
What it reads
Who ranks and who is cited for each question today, set against the demand your own data shows.
What it cannot see
The true search volume. Third-party figures are models built from clickstream panels.
What it leaves behind
A question map: cluster, intent, current holder and the gap that makes a page worth writing.
Technologies we work with
SemrushSearch, SEO & analytics
GoogleSearch, SEO & analytics
04Structured data
Can search engines and AI tell what this page is about?
Schema.org JSON-LD, validated before release and on every build, then monitored in Search Console for errors after launch.
Measured: markup either validates and qualifies, or it does not
Also in this jobSchema.org vocabularyRich Results TestSchema Markup Validator
What it reads
Whether the markup parses and qualifies for rich results, and whether it still does after each release.
What it cannot see
Whether Google will show the rich result.
What it leaves behind
JSON-LD in your templates, a build check that fails on invalid markup and enhancement reports reviewed weekly.
Technologies we work with
Google Search ConsoleSearch, SEO & analytics
WordPressCMS & content
ContentfulCMS & content
05AI visibility
Are we in the answer, and with a link?
A fixed set of buyer prompts, run on a schedule in each AI assistant from a clean session.
Sampled: answers vary from run to run, so results are rates with a sample size
Also in this jobOpenAIPlaywrightOur own prompt-testing toolCommercial AI visibility trackers
What it reads
For each prompt and assistant: cited with a link, mentioned without one or absent, and which sources were cited instead.
What it cannot see
Why the model chose those sources. We record what it did and never claim to know why.
What it leaves behind
A versioned prompt set, the raw answers with timestamps and a weekly rate with its sample size.
Technologies we work with
PerplexityModel providers
Google GeminiModel providers
AnthropicModel providers
PythonLanguages & runtimes
06Reporting
Do all teams read the same numbers?
Every source loads into one data warehouse on a schedule and feeds a single dashboard.
Both, labelled: measured and sampled figures never share a column
Also in this jobScheduled Search Console bulk exportMonthly review document
What it reads
Every source on one schedule, modelled once, so last month reads the same in every meeting.
What it cannot see
Causation. The dashboard shows what moved and when we released work; the written report argues the link.
What it leaves behind
One warehouse dataset, one dashboard and a monthly written report both teams sign off.
Technologies we work with
Google BigQueryData, analytics & streaming
LookerData, analytics & streaming
Google AnalyticsSearch, SEO & analytics
PythonLanguages & runtimes
Why the labels matter. Keyword volumes from third-party tools are estimates, AI visibility figures are samples and Search Console clicks are measurements. We use all three and always say which is which.
08 · MeasurementReporting
One dashboard for rankings and citations.
Measuredthe raw export, before anyone rounds itOne report every figure labelled by how it was producedPhoto: Mika Baumeister / Unsplash
See what search sent you and what AI answers said about you, side by side, in one report fed by six sources.
Visibility dashboard · yourcompany.com · 26 weeks to 13 SepIllustrative
Organic clicks and impressions
Measured · Search Console
4,800
3,600
2,400
1,200
0
240k
180k
120k
60k
0
Mar
Apr
May
Jun
Jul
Aug
Sep
Clicks
Impressions
Week ending 13 Sep ·
4,610 clicks ·
214,300 impressions ·
+7.5% on the week
An illustrative 26-week chart for the placeholder company. Weekly organic clicks rise from 1,840 in the week ending 22 Mar to 4,610 in the week ending 13 Sep, about 151 per cent higher, with impressions rising from 96,200 to 214,300 over the same period. Both lines dip slightly in three separate weeks rather than climbing evenly.
62%
Non-brand share of clicks
Searches for your brand name are excluded, so the figure shows demand you did not already own.
Measured
31%
Citation rate in AI answers
Answers that named you with a working link, over 960 sampled answers.
Sampled
148
Organic-assisted conversions
Conversions where an organic search visit came before the final interaction.
Measured
92%
URLs passing Core Web Vitals
Field data at the 75th percentile over a rolling 28 days.
Measured
Share of answer by engine
Sampled · 12 prompts × 20 runs
Perplexity41% cited · 9% mentioned
Google AI Overview34% cited · 12% mentioned
ChatGPT search29% cited · 16% mentioned
Gemini app22% cited · 18% mentioned
Cited with a link
Mentioned, no link
Absent
Illustrative share of answer for four answer surfaces, each from 120 sampled answers. Perplexity cited the placeholder company in 41 per cent of answers and mentioned it without a link in a further 9 per cent; Google AI Overview 34 and 12 per cent; ChatGPT search 29 and 16 per cent; the Gemini app 22 and 18 per cent. The remainder of each bar is absence.
Data sources
Six sources in one data warehouse
The six sources behind the dashboard: what each provides, how often it refreshes and whether its figures are measured on your own sites, sampled from repeated runs or estimated by a third-party model.
Source
What it gives
Refresh
Kind
Search Console bulk export
Clicks, impressions and position by query, page, country and device
Daily
Measured
Analytics property
Sessions, engaged sessions, conversions and assisted paths
Daily
Measured
Field performance data
LCP, INP and CLS at the 75th percentile of visits, by template
Monthly
Measured
Prompt-testing tool
Cited, mentioned without a link or absent, for each prompt and assistant
Weekly
Sampled
Rank tracker
Position for tracked queries, with feature presence
Weekly
Estimated
Crawl and server logs
Indexable URLs, status codes and what search crawlers fetched
Monthly
Measured
How we read it. Share of answer, how often AI answers mention you, is a rate from repeated runs, reported with the prompt set, run count and date. The same prompt can get a different answer an hour later. Clicks are counts and rank-tracker positions are estimates, so we never average the three or present a sampled rate as traffic.
The reporting rhythm. We re-run the prompt set weekly and send a written report each month against the baseline set in week one. Each monthly report names what moved, what did not and what we would stop doing.
09 · GuardrailsQuality standards
Search rules we follow.
Most search damage we are asked to repair started as a shortcut. We work within the published rules, check every page before it goes live and keep the records.
Every page passes these checks
6 checks · before publish
Crawlable and indexable
Returns 200, has a self-referencing canonical, carries no stray noindex and is in the sitemap we submit.
Built to WCAG 2.2 AA
Correct heading order, labelled controls, visible focus and full keyboard access; the structure that helps screen readers also helps machines read the page.
Passes Core Web Vitals
LCP ≤ 2.5 s, INP ≤ 200 ms and CLS ≤ 0.1 at the 75th percentile of visits.
Accurate structured data
Schema.org JSON-LD that describes only what a visitor can see, validated before release and on every build.
Named author and review date
A named person answers for every claim, with sources linked and the date of the last check.
Consent before measurement
Analytics and advertising tags fire only after consent under the GDPR and India’s DPDP Act 2023. Search Console data is unaffected because it needs no tag.
What we will not do
5 refusals · whoever asks
No cloaking
Crawlers and people get the same page. Showing crawlers different content breaks Google’s spam policies and can be detected.
No mass-produced pages
No pages generated in bulk and published without an editor reading each one.
No bought or exchanged links
Anything paid for is marked rel="sponsored" and listed in our report.
No false markup
No invented reviews or ratings, and no FAQ markup for questions the page does not answer.
No ranking or citation guarantees
Search engines and AI providers decide what ranks and what is cited; we commit to the work, the baseline and the reporting.
Published rules
Google Search Essentials
Technical requirements, spam policies and guidance on helpful, people-first content.
The published requirements for appearing in Google Search. Google says its AI features need no extra markup beyond them. They change, so we re-read them and re-check the site.
Schema.org
The shared vocabulary search engines and assistants read for entities, products, articles and FAQs.
We write JSON-LD to this vocabulary. Valid structured data makes a rich result possible, never certain.
Frameworks we build to
Core Web VitalsLoading, interactivity and visual stabilityGood at the 75th percentile of page loads: LCP ≤ 2.5 s, INP ≤ 200 ms, CLS ≤ 0.1.
WCAG 2.2 AAWeb Content Accessibility GuidelinesPerceivable, operable, understandable and robust content, including 2.2 criteria such as focus not obscured, target size and accessible authentication.
GDPRGeneral Data Protection Regulation (EU) 2016/679Lawful basis, data-subject rights, data protection by design and by default, breach notification and DPIAs for high-risk processing.
DPDP Act 2023Digital Personal Data Protection Act, 2023Notice and consent, duties of data fiduciaries, rights of data principals, breach intimation and added duties for significant data fiduciaries, with the DPDP Rules.
SCI · ISO/IEC 21031:2024Software Carbon IntensityA rate of carbon emissions per functional unit: SCI = ((E × I) + M) per R, where E is energy, I is grid carbon intensity, M is embodied emissions and R the unit.
Where a standard offers certification, an accredited body issues it, never us.
Smaller pages, fewer redirects
Three fixes make pages cheaper to crawl and serve and faster for readers: one URL per page, one redirect hop and a page-weight budget.
Redirect chains collapsed
One 301 hop, then stop; every extra hop is a request that shows nothing to anyone.
Duplicate and parameter URLs consolidated
Crawl effort goes to pages people read instead of to eleven addresses for one page.
Page weight budgeted per template
A budget agreed at design time and checked in the build, so a page cannot quietly double in size.
What we can honestly claim. The Software Carbon Intensity specification scores software as SCI = ((E × I) + M) per R: energy times the carbon intensity of the electricity, plus embodied emissions, per unit of work. Fewer wasted crawls and lighter pages lower E and the number of requests. They do not change I or M, so we leave those out of anything we report.
10 · ProcessHow we work with you
Technical fixes first, then content and citations.
A page that crawlers cannot read will not rank on Google or be cited in AI answers. Fixes go live in weekly batches instead of one release at the end, because AI answers change while the work is under way.
Each quarter, the results feed the next audit.
Audit
Wk 01–03
Can search engines and AI read and understand your site today?
Technical crawl, log and index analysis, a content and entity audit, then a baseline of rankings and AI-answer presence for your priority questions.
Every indexable URL crawled as rendered after JavaScript runs, the way a search engine sees it.
A server log sample showing what each crawler requested, how often and with what status code.
Index coverage, structured data errors and Core Web Vitals field data read from Search Console instead of one lab test.
A prompt set built from your buyers’ questions, run repeatedly in each AI assistant to give a baseline rate with a sample size.
You keep
Technical audit
Visibility baseline
Prompt set
What we need from you
Read access to Search Console and analytics, a server log sample and a staging URL if you have one.
Done when
The baseline is signed off and the fix backlog ranked. Nothing changes before the baseline is recorded.
Fix
Wk 03–06
What stops your key pages being crawled, rendered, indexed and loaded quickly?
Indexation, rendering, speed, internal linking and structured data fixed with your developers, highest impact first.
Status codes, canonicals, redirect chains and parameter URLs cleaned up, so crawlers spend their time on pages that matter.
Pages that must be indexed are server-rendered or pre-rendered instead of relying on crawlers to run your JavaScript.
Core Web Vitals improved on field data until all three pass at the 75th percentile of visits.
Schema.org JSON-LD validated on every build, so a template change cannot quietly break it months later.
You keep
Fix backlog
Schema implementation
Core Web Vitals fixes
What we need from you
A developer with deploy rights, or our engineers working inside your repository under your review process.
Done when
Priority URLs return 200, render without errors, validate as structured data and pass Core Web Vitals in field data.
Build
Wk 06–12
Which buyer questions should you answer, and who else must mention you?
We publish topic clusters, answer-first pages and expert content, and seek citations from trusted sites through digital PR.
Question clusters chosen from demand shown in query data, sales calls and support tickets.
Pages that answer in the first fifty words, then back it up with a comparison, original data, a named author and sources.
Organization markup, consistent contact details and matching external profiles, built in parallel.
Mentions earned in the trade publications, review sites, documentation and communities AI answers already cite.
You keep
Content cluster
Entity updates
Citation plan
What we need from you
A subject-matter expert for about an hour a week, and approval to publish under a named author.
Done when
The cluster is live with named authors, sources and update dates, and outreach is under way with mentions tracked.
Measure
Monthly
Are the numbers moving, and which work moved them?
Rankings, organic traffic and conversions, AI Overview presence and share of answer across engines, reported against the baseline.
Clicks, impressions and average position exported from Search Console on a schedule.
The prompt set re-run weekly, because one check proves little.
Share of answer, citation rate and non-brand share reported by AI assistant and intent, against the pre-change baseline.
A written report each month: what moved, what we released, what changes next and what we got wrong.
You keep
Monthly report
Share-of-answer tracking
Next-quarter plan
What we need from you
Thirty minutes a month from whoever owns the sales pipeline, so the numbers are read next to revenue.
Done when
Both teams read one dashboard, with measured and sampled figures labelled and never averaged together.
The first ninety days, week by week
W13of 13 weeks
Workstream
Stage
01Week 01
02Week 02
03Week 03
04Week 04
05Week 05
06Week 06
07Week 07
08Week 08
09Week 09
10Week 10
11Week 11
12Week 12
13Week 13
Runs
Crawl, log and index analysis
Rendered-DOM crawl and a log sample by user agent
Audit
Runs week 01 to week 03
W01–W03
Baseline and prompt panel
Rankings, clicks and answer presence recorded before any change
Audit
Runs week 02 to week 03
W02–W03
Indexation and rendering
Status codes, canonicals, internal links, rendering path
Fix
Runs week 03 to week 06
W03–W06
Structured data and entities
JSON-LD written, validated, then watched for enhancement errors
Fix
Runs week 04 to week 06
W04–W06
Answer-first content
Clusters published with named authors, sources and updated dates
Build
Runs week 06 to week 12
W06–W12
Digital PR and citations
Commentary, data and review programmes where answers already look
Build
Runs week 07 to week 12
W07–W12
Dashboard and reporting
Weekly re-sampling, one dashboard, a monthly read with both teams
Measure
Runs week 04 to week 13Monthly report
Monthly report
Monthly report
W04–W13
Illustrative schedule. Seven workstreams across thirteen weeks; the reporting row carries a monthly read at weeks 04, 08 and 12. Work continues on a monthly cycle after week 13.
The strip fills from week one to week thirteen when it comes into view; the finished state shows every workstream in place.
Audit
Fix
Build
Measure
Monthly report
Why weekly, not quarterly. A technical fix can show within days of a recrawl, while an AI answer can change with nobody publishing anything. The monthly report explains the trend instead of one lucky run.
11 · DeliverablesWhat you get
Six deliverables your team can run.
Each deliverable has an owner, a first delivery date and an update schedule. You receive tickets, markup, sheets, pages and dashboards, ready to use after handover.
01
Technical SEO audit & prioritised backlog
Report · tickets
A rendered crawl of every indexable URL, a server-log sample split by crawler and index coverage from Search Console. Fixes are ranked by traffic at risk and effort, then filed as developer-ready tickets in your tracker.
First delivered
Week 03
Kept current
Re-crawled monthly, re-audited each quarter
02
Structured data implementation
JSON-LD
Schema.org markup for Organization, Product, Article with a named author, BreadcrumbList and FAQPage only where a page qualifies. Validated before release and in every build, then monitored in Search Console for errors.
First delivered
Week 06
Kept current
Validated on every release
03
Topic & entity map
Sheet · board
Your company facts (legal name, locations, contacts, products and people), checked against the external profiles that repeat them and mapped to the buyer questions that lead to revenue. It guides both the content plan and the markup.
First delivered
Week 04
Kept current
Reviewed monthly
04
Answer-first content
CMS pages
Pages that answer in the first fifty words, with a comparison table, original data, a named author with credentials, an accurate last-updated date and sources. Agents draft briefs and check claims against sources; an editor writes, approves and signs each page.
First delivered
Week 06 onwards
Kept current
Published to a calendar, refreshed on a schedule
05
AI visibility prompt set & tracker
Dashboard
The prompt set, the code that runs it and the run history: cited, mentioned without a link or absent, by engine and prompt. Reported as rates with sample sizes, since AI answers vary between runs.
First delivered
Week 03
Kept current
Re-sampled weekly
06
Monthly visibility report
Dashboard · PDF
Search Console clicks, impressions and position beside share of answer and citation rate by engine, plus organic-assisted conversions and Core Web Vitals pass rate. Measured and sampled figures are labelled and never averaged together.
First delivered
Month 01
Kept current
Monthly, against the baseline
Services & packages
Be found on Google and cited in AI answers.
We make your brand easy to find in Google and Bing, and easy for ChatGPT, Perplexity, Gemini and Google’s AI Overviews to cite. Technical SEO, content and AI visibility run as one programme, reported against a baseline.
01Pick services. Enquire about one, or add several to a brief.
02Choose a package. A sprint, a fixed project or an ongoing team.
03Send the brief. We reply within one working day.
How we work with you
Ways to engage, from a question to an RFQ.
Ask a quick question, send a project brief or issue a formal RFQ. The lead for the work reads each one in full, and any services already in your brief go with it.
01The questionsThe prompts and searches your buyers use, and who is cited for them today.
02The evidenceRead access to Search Console and analytics, so the baseline is measured rather than guessed.
03The blockersA sample of server logs and a staging URL, to see where a crawler stops and what that costs.
04The shapeWhat a realistic first ninety days looks like against your release process.
Roughly forty-five minutes. No deck, and no proposal until we have seen the data.
SEO is for rankings, AEO for direct answers and GEO for AI assistants. AEO (answer engine optimisation) shapes content for featured snippets and AI Overviews. GEO (generative engine optimisation) helps ChatGPT, Perplexity and Gemini understand and cite your brand. All three share the same foundations, so we run them as one programme.
No. AI answers draw on pages that are crawlable, well structured and cited by trusted sources, which is what good SEO produces.
Decide per bot. Googlebot (Search and AI Overviews), OAI-SearchBot (ChatGPT search) and PerplexityBot fetch pages for search; block them and you disappear from those answers. GPTBot collects OpenAI training data and Google-Extended controls how Gemini uses your content, without affecting Search rankings; decide those with your legal team. robots.txt is a request, so keep private content behind a login.
No major engine has confirmed it affects ranking or citation. llms.txt is a community proposal: a Markdown file listing the pages you want AI models to read. It costs little, so we add it where it can stay accurate and track the effect. Nothing depends on it: citations come from crawlable pages, clear structure, correct company facts and earned mentions.
We regularly run a set of buyer prompts through the main AI engines. Each run records whether your brand is mentioned, cited or linked, and we report how that share changes over time.
No. Nobody can guarantee either, because search engines and AI providers control their own results. We commit to the agreed work, the measurement and reporting against a baseline.
Technical fixes can show within weeks of being recrawled. Content and authority typically take three to six months, depending on competition and how often you publish.
Search and AI results change quickly. When an answer here stops being true, we rewrite it, as we do with your pages.
