Pick the best model for each task, and keep the freedom to switch vendors.
Gemini 3.5 Pro just slipped to July and Google lost four senior Gemini and AlphaFold researchers in two weeks. Perplexity runs across 19 models including Claude and GPT, so each task goes to the model that handles it best.
Best-of-breed model choice in one product. Google can only give you Gemini.
Perplexity's Computer orchestrates 19 models (Claude, Gemini, GPT, Grok) and routes each subtask to the strongest one; architects who don't want to bet their whole AI stack on a single vendor get real optionality that a Gemini-only Google cannot offer.
Ad-free, cited answers a buyer can defend. Google's answers are ad-supported and increasingly wrong at scale.
Perplexity dropped advertising entirely in 2026 and shows inline sources on every answer; for economic buyers in research, finance, legal and media who must justify what they act on, that verifiability and absence of commercial bias is the whole point.
Vendor-neutral AI cleared for government: a wedge where single-model pitches stall.
Perplexity Enterprise Pro for Government earned FedRAMP prioritization and a GSA OneGov deal as the 'first multi-model AI platform,' letting federal teams reach Gemini, Claude, GPT and Sonar in one accredited platform without locking into one vendor, exactly what risk-averse public-sector and regulated buyers want.
Expand section · 8 moreCollapse section▾
UpdatedOn August 13, 2026 Google launched Gemini 3.7 Flash, its workhorse model for coding and agents, just three weeks after Gemini 3.6 Flash, at an introductory price of $0.75 per million input tokens and $3.75 per million output tokens (half the prior model's launch price). Coding and agent benchmarks jumped over 3.6 Flash: DeepSWE v1.1 65.3% vs 49.0% and WebDev Arena Elo 1588 vs 1538.
Expand section · 15 moreCollapse section▾
UpdatedGoogle is discontinuing the legacy Google Assistant on Android phones, tablets and paired devices (Wear OS watches, headphones, Android Auto projection) starting September 4, 2026, forcibly replacing it with Gemini. Once a device transitions there is no way to switch back, and Google is notifying users by email. Vehicles with Google Built-in keep Assistant for now.
On August 5, 2026 the UK Competition Appeal Tribunal certified an opt-out collective action against Google covering roughly 880,000 UK-domiciled advertisers and seeking up to £5 billion (about $6.7 billion). The claim alleges Google abused its dominance across mobile operating systems, app distribution and search advertising (tying Search and Chrome onto Android, paying Apple for default placement, and biasing its Search Ads 360 platform toward Google) to overcharge advertisers. Certification is the procedural gate to a trial on the merits, not a finding that Google broke the law; Google called it 'speculative and opportunistic' and will contest it.
On July 23, 2026 the European Commission fined Google €890 million ($1 billion) in its first and largest-ever DMA penalty: €460 million for self-preferencing its own shopping, hotels, transport and sports results in Google Search over rivals, and €430 million for blocking Google Play developers from steering users to cheaper channels outside the store. Google has 60 days to end both practices or risk periodic penalties up to 5% of its worldwide turnover, and says it is reviewing the decision and weighing an appeal.
On July 16, 2026 the European Commission adopted a binding DMA specification decision under Article 6(11) requiring Google to share anonymised Search ranking, query, click and view data with rival search engines, and it explicitly makes AI chatbots that offer search functionality eligible recipients of that same data. Sharing runs on FRAND cost-recovery pricing and is due to begin January 2027. Eligibility is limited to a genuine EU search engine running at least two consecutive years, or one founded within two years that raised over €50 million and has at least 50,000 monthly EU users. The measure carries no financial penalty; it specifies how Google must comply after regulators found its earlier data-sharing offer, which excluded AI chatbots and stripped out 90 to 100% of unique queries, produced no meaningful uptake.
Google's Gemini 3.5 Pro is months behind schedule, not shipped as the leaked July 17 target claimed. Bloomberg reported on July 16, 2026 that the flagship model's coding capabilities were short of internal expectations, and Alphabet shares fell 4% as OpenAI and Meta shipped stronger coding models.
A July 8, 2026 EU General Court ruling closed Google's last route to pre-empt two binding DMA decisions due July 27, 2026. One (case DMA.100220, Article 6(7)) forces Google to give rival AI services free and effective interoperability with the Android system features Gemini uses: wake-word invocation, the long-press gesture, on-screen context awareness and neural-processing-unit access. The other (DMA.100209, Article 6(11)) forces Google to share anonymized Search ranking, query and click data with rival search engines and AI chatbot providers on FRAND terms. Compliance begins immediately on issuance; Google can challenge only the content afterward, and non-compliance fines reach up to 10% of Alphabet turnover (over $35 billion).
On July 1, 2026 the Patent and Market Court in Stockholm ordered Google to pay about $1.97 billion in damages plus accrued interest to Klarna's PriceRunner for illegally favoring its own comparison shopping service for years, the largest competition damages award in Swedish history. The case is a follow-on to the European Commission's 2017 Google Shopping decision, and Google says it disagrees with the ruling and is weighing an appeal, so a final payout could be years away.
Google's AI talent drain went from a single marquee exit to a sustained run. Inside a fortnight, Gemini co-author and 'Attention Is All You Need' author Noam Shazeer left for OpenAI, then DeepMind Nobel laureate John Jumper, AI-coding lead Jonas Adler, and pretraining specialist Alexander Pritzel all left for Anthropic. Four senior departures from the team behind Gemini and AlphaFold in a matter of days, three of them to the same rival, Anthropic. Bloomberg's reporting cites compute access (researchers queuing for Google's own TPUs) and pre-IPO equity at Anthropic and OpenAI as the pull.
On June 18, 2026 Perplexity launched Brain, a self-improving memory system for its Computer agent, in Research Preview for Max ($200/month) and Enterprise Max subscribers. Brain builds a traceable context graph of what the agent did, what worked, and what got corrected, then synthesizes it overnight into a per-user LLM wiki loaded before each task. Perplexity's internal tests report 25% higher answer correctness on repeated tasks, 16% better recall, and 13% lower cost on context-heavy work. The figures are first-party with no independent benchmark yet.
Perplexity is now a disciplined pre-IPO company, not a flailing startup: it's holding a self-imposed 2028 IPO line by choice while rivals rush to file. CEO Aravind Srinivas told CNBC on June 9, 2026 that Perplexity plans to go public in 2028 regardless of how the confidential IPO filings of Anthropic (valued near $1T) and OpenAI ($852B) perform; Chief Business Officer Dmitry Shevelenko framed holding 2028 as the earliest date as a strategy that let Perplexity 'build a healthy, high-growth business.'
Within two weeks of Google's May 19 AI-search overhaul, DuckDuckGo traffic ran ~84% above baseline and its no-AI search page visits tripled on May 28, a measurable user-defection signal against Google force-feeding AI with no opt-out.
At Google I/O on May 19, 2026, Google made Gemini 3.5 Flash the default model in AI Mode for everyone globally and shipped 'the biggest upgrade to our Search box in over 25 years,' plus background 24/7 Search Agents, directly adopting Perplexity's answer-and-agent framing at Google's scale.
At I/O (May 19, 2026) Google restructured consumer AI subscriptions, launching a $100/month AI Ultra tier and cutting its top Ultra plan from $250 to $200, a signal Google was mispriced and is now competing aggressively on AI value, not just distribution.
An Ahrefs study found Google AI Overviews correlate with a 58% drop in click-throughs for top-ranking pages (nearly double the prior year's decline), triggering a Penske Media antitrust suit and an EU competition investigation; a narrative lever for Perplexity in media/publishing verticals.
On February 3, 2026, the DOJ and a coalition of states appealed the search-monopoly remedies ruling, calling the behavioral remedies a 'slap on the wrist for a recidivist monopolist' and seeking stronger structural relief, keeping Google's default-distribution model legally unsettled into 2026 and 2027.
Perplexity positions itself as an 'answer engine' replacing ranked blue links with a single synthesized, cited answer, and frames accuracy as foundational. CEO Aravind Srinivas: 'hallucination is a bug,' not a feature (the company's own positioning).
Expand section · 3 moreCollapse section▾
As of June 2026 Perplexity has repositioned from answer engine to a 'hybrid neutral orchestration layer': a model- and chip-agnostic platform that routes each task to the best model across device and cloud, which it frames as its durable advantage versus Google, OpenAI and Anthropic (company claim).
Perplexity abandoned advertising entirely in early 2026, with an executive arguing 'a user needs to believe this is the best possible answer', positioning trust and freedom from commercial bias against Google's structurally ad-funded search (company claim).
Google positions itself as a full-stack, billion-user AI company (custom silicon to models to products touching billions) and now openly adopts the 'answer' and 'agent' framing Perplexity pioneered, but backed by distribution Perplexity cannot match (company claim).
Perplexity's tiers run Free, Pro at $20/month, and Max at $200/month; the Comet agentic browser launched as a $200/month Max-only benefit before going free worldwide in October 2025.
Expand section · 4 moreCollapse section▾
Perplexity's Sonar API is priced per million tokens (Sonar $1 in/$1 out; Sonar Pro $3 in/$15 out; Deep Research $2 in/$8 out) plus per-request and citation/search fees, a usage-based developer model competing with Google's Gemini API/Vertex.
Google AI Pro is $19.99/month, at functional price parity with Perplexity Pro ($20), but bundles YouTube Premium Lite, 5TB storage, and Gemini across Gmail/Docs, none of which Perplexity offers.
Google no longer sells Gemini as a standalone Workspace add-on: it is bundled into Business Standard ($14/user/month) and above, meaning Google's enterprise AI distribution is effectively zero incremental cost to existing Workspace customers.
Perplexity leans on partner bundles to reach users outside its paywall: Deutsche Telekom's T Phone 3, for example, includes 18 months of Perplexity Pro ($360 value) free. It's a distribution workaround against Google's native free reach, but one that creates churn risk when promos expire.
Competitive Battlecard8 across 3 zones›
Best-of-breed model choice in one product. Google can only give you Gemini.
Best for Technical evaluatorPerplexity's Computer orchestrates 19 models (Claude, Gemini, GPT, Grok) and routes each subtask to the strongest one; architects who don't want to bet their whole AI stack on a single vendor get real optionality that a Gemini-only Google cannot offer.
Expand · 2 moreCollapse▾
Ad-free, cited answers a buyer can defend. Google's answers are ad-supported and increasingly wrong at scale.
Best for Economic buyerPerplexity dropped advertising entirely in 2026 and shows inline sources on every answer; for economic buyers in research, finance, legal and media who must justify what they act on, that verifiability and absence of commercial bias is the whole point.
Vendor-neutral AI cleared for government: a wedge where single-model pitches stall.
Best for Security & regulatedPerplexity Enterprise Pro for Government earned FedRAMP prioritization and a GSA OneGov deal as the 'first multi-model AI platform,' letting federal teams reach Gemini, Claude, GPT and Sonar in one accredited platform without locking into one vendor, exactly what risk-averse public-sector and regulated buyers want.
The agentic race is genuinely contested: both are shipping autonomous agents right now.
Best for Eng-led championGoogle launched 24/7 background Search Agents at I/O while Perplexity shipped Computer for Enterprise with 100+ connectors; neither has a decisive deployment lead, so deals turn on integration fit and trust, not on who has 'agents' at all.
Expand · 1 moreCollapse▾
At the Pro tier, price is a wash, so don't let the deal become a price contest.
Best for Economic buyerPerplexity Pro ($20) and Google AI Pro ($19.99) are effectively identical on price, but Google folds in YouTube Premium and Workspace. On sticker price alone Perplexity loses the bundle math, which is exactly why these deals must be steered to neutrality, citations and connectors.
Google's distribution is overwhelming: its AI is the default for over a billion people at zero acquisition cost.
Best for Exec / top-downAI Mode hit 1B monthly users free inside Search, Chrome and Android in a single year; any Perplexity buyer has to choose to leave a tool already in everyone's hands. Be honest that this is real, then move the conversation to where defaults don't decide: accredited, neutral, cited enterprise workflows.
Expand · 2 moreCollapse▾
Gemini 3 is a genuinely top-tier model: concede the benchmark fight and win the trust fight.
Best for Technical evaluatorGemini 3 Pro topped LMArena at 1501 Elo with PhD-level reasoning scores; Perplexity builds no frontier model of its own. For benchmark-driven technical buyers this is Google's strongest card, so pivot from raw model scores to answer accuracy, citations and the freedom to use Gemini inside Perplexity when it's the best fit.
Workspace lock-in is real: Gemini lives inside the Gmail, Docs and Meet that 3B people already use.
Best for Economic buyerGoogle embeds AI into the daily apps of over 3 billion Workspace users, so for an org already standardized on Google, switching cost is organizational, not technical. Don't fight replacement. Sell Perplexity as the additive, neutral research layer alongside Workspace.
Sentiment5 signal›
Perplexity users repeatedly praise its citation-backed answers and lower hallucination rate as the main reason they trust it for research over standard chatbots (G2 review sentiment).
A recurring user complaint is that Perplexity's Deep Research can be confidently wrong and fail to flag uncertainty, undermining the very source-grounding that is its selling point (Reddit sentiment).
A Guardian investigation found Google AI Overviews giving health advice doctors called 'really dangerous' (e.g. telling pancreatic cancer patients to avoid high-fat foods), fueling a trust narrative Perplexity can use in high-stakes domains.
Publishers and users describe Google's AI Mode turning search into a zero-click engine (~60% of queries, 69% for news) with one outlet calling it an 'extinction-level event', driving defections to DuckDuckGo, Kagi and Brave (documented backlash).
Paid Gemini subscribers revolted in late May 2026 when a new compute-based quota let a single prompt exhaust a five-hour window, forcing Google to publicly revise the limits within days (user sentiment).
Objection Handling7 objections›
"Google gives me AI answers free in Search. Why would I pay $20 for Perplexity?"
Raised by Economic buyerIt's a fair question: Google AI Mode is genuinely free to over a billion users. But Google's free answers are ad-supported and single-model, and at Google's scale independent analysis pegs AI Overviews at tens of millions of questionable answers an hour. The $20 buys an ad-free, source-cited answer and the ability to route each question to the best of 19 models: value, not just access.
"Perplexity is just a wrapper on other people's models. Google owns its models and infrastructure."
Raised by Technical evaluatorThat critique was fair early on, and TechCrunch has named it. The honest counter: Perplexity now runs its own AI-optimized web index and Sonar models and no longer depends on others' APIs for search, and the real IP is the orchestration layer that picks the best of 19 models per task in real time. Owning one model is a constraint; orchestrating all of them is the product.
"Nine publishers are suing Perplexity for copyright. What's our exposure if we standardize on it?"
Raised by Security & regulatedThis is real and unresolved: CNN, the NYT, News Corp and six others have active suits as of May 2026. The honest answer is that Perplexity is simultaneously signing licensing deals (Gannett, TIME, Le Monde, Der Spiegel), its fair-use 'you can't copyright facts' defense has genuine legal footing, and this is a category-wide AI risk Google itself has faced and settled via licensing, not a Perplexity-specific defect.
"I read Perplexity's Comet browser had serious security holes. Can we trust it with our data?"
Raised by Security & regulatedDon't dodge it: security researchers disclosed prompt-injection and data-exfiltration flaws in Comet, and Perplexity initially under-rated one report before patching by February 2026. Since then Perplexity has rebuilt the runtime underneath its agents. SPACE now runs 100% of Computer agent sessions, isolating every task in its own virtual machine with per-session credential isolation, and credentials are injected from outside the sandbox rather than stored inside it. Indirect prompt injection remains an industry-wide agentic-browser problem (Brave and others have flagged the same class), but Perplexity has paired the patched vectors with this hardened isolation model and its FedRAMP-grade controls for sensitive deployments.
"Is Perplexity stable enough to bet on against a trillion-dollar Google?"
Raised by Economic buyerIt's a legitimate concern: at a $20B valuation Perplexity is a fraction of Alphabet's ~$4.5T. But it tripled annualized revenue in 2026 to roughly $500M ARR, raised ~$1.5B from SoftBank, Nvidia and others, and counts the bulk of the Fortune 500 as users. The question that matters is whether the product survives and delivers ROI over the buyer's horizon, and the trajectory says yes.
"I heard a lawsuit says Perplexity sends user chats to Google and Meta, even in Incognito."
Raised by Security & regulatedA March 2026 class action does allege embedded trackers forwarded chat data to Google and Meta without consent. Reps will be asked about it, so don't bury it. The honest answer: these are unproven allegations Perplexity is contesting, paid Pro/Max subscribers are excluded from the proposed class, and Perplexity's enterprise and FedRAMP-cleared government offerings run under separate, accredited data controls from the consumer web product.
"Perplexity quietly cut Pro usage limits. Is the pricing even stable?"
Raised by Economic buyerUsers did complain on Reddit about tighter rate limits in 2026, and it's a fair flag. Perplexity attributes the change to cracking down on promo-code resale fraud rather than degrading the product, and an executive publicly denied the free tier was being intentionally throttled. For business buyers the relevant tier is Enterprise, which carries its own SLAs separate from consumer Pro limits.
Cut Log8 removed / revised›
as_of = the date the fact is true as-of · verified_on = when grounding last confirmed the exact wording · is_new = a monitor run touched it <48h ago.| subject_key | section | as_of | verified_on | is_new |
|---|---|---|---|---|
| perplexity-vs-google | competitive-frame | 2026 | executive_summary | 2026-06-24 | 2026-06-06 | false |
| perplexity | legal-exposure | 2026 | executive_summary | 2026-08-04 | — | false |
| google | distribution-under-legal-attack | 2026 | executive_summary | 2025-09-02 | 2026-06-06 | false |
| perplexity-vs-google | accuracy-trust-wedge | 2026 | executive_summary | 2026-04-08 | 2026-06-06 | false |
| perplexity-vs-google | price-vs-value | 2026 | executive_summary | 2026-05-19 | 2026-06-06 | false |
| alphabet | q-revenue | q1-2026 | snapshot | 2026-04-29 | 2026-06-06 | false |
| google | search-revenue | q1-2026 | snapshot | 2026-04-29 | 2026-06-06 | false |
| google-cloud | q-revenue | q1-2026 | snapshot | 2026-04-29 | 2026-06-06 | false |
| alphabet | valuation | current | snapshot | 2026-06-05 | 2026-06-06 | false |
| google | ai-mode-mau | current | snapshot | 2026-05-19 | 2026-06-06 | false |
| google | ai-overviews-reach | current | snapshot | 2026-05-19 | 2026-06-06 | false |
| google | gemini-app-mau | current | snapshot | 2026-05-19 | 2026-06-06 | false |
| perplexity | revenue-run-rate | current | snapshot | 2026-04-15 | 2026-06-06 | false |
| perplexity | valuation | current | snapshot | 2026-06-03 | 2026-06-06 | false |
| perplexity | active-users | current | snapshot | 2026-04-08 | 2026-06-06 | false |
| google | search-backlash | duckduckgo-surge | recent_moves | 2026-06-01 | 2026-06-06 | false |
| google | legal-action | eu-dma-fine | recent_moves | 2026-07-23 | 2026-07-24 | false |
| google | launch | ai-mode-io-2026 | recent_moves | 2026-05-19 | 2026-06-06 | false |
| google | pricing-change | ai-subscription-overhaul | recent_moves | 2026-05-19 | 2026-06-06 | false |
| google | ai-overviews-traffic-impact | 2026 | recent_moves | 2026-05-06 | 2026-06-06 | false |
| google | legal-action | doj-remedies-appeal | recent_moves | 2026-02-03 | 2026-06-06 | false |
| perplexity | positioning | answer-engine | positioning | 2025-07-24 | 2026-06-06 | false |
| perplexity | positioning | orchestration-layer | positioning | 2026-06-03 | 2026-06-06 | false |
| perplexity | positioning | ad-free | positioning | 2026-02-18 | 2026-06-06 | false |
| google | positioning | full-stack-agentic | positioning | 2026-05-19 | 2026-06-06 | false |
| perplexity | list-price | max-tier | pricing | 2025-10-02 | 2026-06-06 | false |
| perplexity | list-price | sonar-api | pricing | 2026-06-06 | 2026-06-06 | false |
| google | list-price | ai-pro | pricing | 2026-05-19 | 2026-06-06 | false |
| google | pricing-model | workspace-gemini-bundled | pricing | 2026-06-06 | 2026-06-06 | false |
| perplexity | pricing-model | telco-bundle | pricing | 2025-08-15 | 2026-06-06 | false |
| perplexity | battlecard | multi-model-orchestration | battlecard | 2026-06-02 | 2026-06-06 | false |
| perplexity | battlecard | ad-free-cited | battlecard | 2026-02-18 | 2026-06-06 | false |
| perplexity | battlecard | fedramp-gov-neutrality | battlecard | 2025-11-19 | 2026-06-06 | false |
| perplexity-vs-google | battlecard | agentic-race | battlecard | 2026-05-19 | 2026-06-06 | false |
| perplexity-vs-google | battlecard | price-parity | battlecard | 2026-05-19 | 2026-06-06 | false |
| google | battlecard | distribution-moat | battlecard | 2026-05-19 | 2026-06-06 | false |
| google | battlecard | gemini-benchmark-strength | battlecard | 2025-11-18 | 2026-06-06 | false |
| google | battlecard | workspace-lockin | battlecard | 2026-04-22 | 2026-06-06 | false |
| perplexity | sentiment | citations-trust | sentiment | 2026-02-10 | 2026-06-06 | false |
| perplexity | sentiment | deep-research-confidently-wrong | sentiment | 2026-03-10 | 2026-06-06 | false |
| google | sentiment | ai-overviews-dangerous-health | sentiment | 2026-01-05 | 2026-06-06 | false |
| google | sentiment | zero-click-publisher-backlash | sentiment | 2026-05-21 | 2026-06-06 | false |
| google | sentiment | gemini-usage-limit-revolt | sentiment | 2026-05-29 | 2026-06-06 | false |
| objection | google-free-vs-paid | objection_handling | 2026-05-19 | 2026-06-06 | false |
| objection | perplexity-is-a-wrapper | objection_handling | 2026-02-27 | 2026-06-06 | false |
| objection | perplexity-copyright-lawsuits | objection_handling | 2026-05-28 | 2026-06-06 | false |
| objection | comet-security-vulnerabilities | objection_handling | 2026-07-15 | — | false |
| objection | perplexity-financial-stability | objection_handling | 2026-06-03 | 2026-06-06 | false |
| objection | perplexity-privacy-data-sharing | objection_handling | 2026-04-01 | 2026-06-06 | false |
| objection | perplexity-usage-limit-cuts | objection_handling | 2026-02-27 | 2026-06-06 | false |
| perplexity | strategic-milestone | ipo-roadmap | recent_moves | 2026-06-09 | 2026-06-10 | false |
| perplexity | product | brain-agent-memory | recent_moves | 2026-06-18 | 2026-06-19 | false |
| google | talent-loss | deepmind-defections | recent_moves | 2026-06-25 | 2026-06-27 | false |
| google | product-delay | gemini-35-pro | recent_moves | 2026-07-16 | 2026-07-17 | false |
| google | legal-action | klarna-pricerunner-damages | recent_moves | 2026-07-01 | 2026-07-02 | false |
| google | legal-action | eu-dma-android-ai-interop | recent_moves | 2026-07-09 | 2026-07-10 | false |
| perplexity | product | space-sandbox | tracked_facts | 2026-07-15 | 2026-07-16 | false |
| google | legal-action | eu-dma-search-data-sharing | recent_moves | 2026-07-16 | 2026-07-28 | false |
| perplexity | legal-action | comet-amazon-injunction | tracked_facts | 2026-08-04 | 2026-08-05 | false |
| google | legal-action | uk-search-ads | recent_moves | 2026-08-05 | 2026-08-08 | false |
| google | product-launch | gemini-37-flash | recent_moves | 2026-08-13 | 2026-08-18 | true |
| google | product-discontinuation | assistant-to-gemini-migration | recent_moves | 2026-08-06 | 2026-08-18 | true |